A Novel Popular Tourist Attraction Discovering Approach Based on Geo-Tagged Social Media Big Data

Size: px
Start display at page:

Download "A Novel Popular Tourist Attraction Discovering Approach Based on Geo-Tagged Social Media Big Data"

Transcription

1 International Journal Geo-Information Article A Novel Popular Tourist Attraction Discovering Approach Based on Geo-Tagged Social Media Big Data Xia Peng 1,2 and Zhou Huang 3, * 1 Collaborative Innovation Center etourism, Institute Tourism, Beijing Union University, Beijing , China; ivy_px@163.com 2 State Key Laborary Resources and Environmental Information System, Institute Geographical Sciences and Natural Resources Research, Chinese Academy Sciences, Beijing , China 3 Institute Remote Sensing & GIS, Peking University, Beijing , China * Correspondence: huangzhou@pku.edu.cn; Tel.: Academic Edirs: Sisi Zlatanova, Jamal Jokar Arsanjani and Wolfgang Kainz Received: 5 May 2017; Accepted: 7 July 2017; Published: 13 July 2017 Abstract: In big data era, social media data that contain users geographical locations growing explosively. These kinds spatiotemporal data provide a new perspective for us observe human movement behavior. By mining such spatiotemporal data, we can incorporate users collective wisdom, build novel services and bring convenience people. Through spatial clustering original user locations, both natural boundaries and human activity information attractions generated, which facilitate performing popularity analysis attractions and extracting travelers spatio-temporal patterns or travel laws. On one hand, potential extracted knowledge could provide decision supports urism management department in both urism planning and resource development; on or hand, travel preferences able be extracted from clustering-generated attractions, and thus, intelligent urism recommendation services could be developed for promote realization smart urism. Hence, this paper proposes a new method for discovering popular attractions, which extracts hotspots through integrating spatial clustering and text mining approaches. We carry out attraction discovery experiments based on Flickr geotagged images within urban a Beijing from The results show that compd with traditional DBSCAN method, this novel approach can distinguish adjacent high-density as when discovering popular attractions and has better adaptability in case an uneven density distribution. In addition, based on finding results scenic hotspots, this paper analyzes popularity distribution laws Beijing s attractions under different temporal and wear contexts. Keywords: social media; geographical big data; attraction; popularity analysis 1. Introduction In era big data, with development mobile Internet technology and popularity intelligent mobile terminals, people increasingly accusmed obtaining or sharing information through mobile intelligent terminal applications whenever and wherever possible. Among numerous mobile applications for information acquisition and sharing, Location-Based Service (LBS) has become mainstream. In process using such applications, massive amounts social media data containing geographic location information (i.e., geo-tagged social media big data) have been generated; moreover, volume such data is exploding. The emergence this new type massive social media data has brought new opportunities and challenges many research fields, attracting researchers interests and attentions. ISPRS Int. J. Geo-Inf. 2017, 6, 216; doi: /ijgi

2 ISPRS Int. J. Geo-Inf. 2017, 6, For example, on Chinese mainland, through Sina micro-blog, user can attach his or her location published text or picture. Similarly, many applications providing local life information services (such as Dianping.com, Baidu NearYou, Jiepang.com, etc.) allow people check-in in restaurants, hotels, attractions and or businesses and comment on business s products or services. Some pho-sharing applications (e.g., Flickr and Instagram) permit users add not only textual description, but also ir current location picture y take, hence called geo-tagged phos. A number travel experience-recording and sharing-applications (such as Baidu urism, Bread Trip and so forth) enable users record ir travel trajecry, take pictures and write travel notes anytime on ir trip. These aforementioned social media data (such as geo-tagged phos, check-in data) contain not only description information, such as title, tag and so forth, but also time information time phographing or checking and spatial location information latitude and longitude place where user ok pho or checked in. A typical application domain geo-tagged data mining is discovery popular attractions. As attractions ten frequently phographed and n uploaded on social media platform, related research on finding popular attractions and recommending appropriate attractions that match users interest has quickly become a hotspot [1 5]. The geo-tagged pho collections or check-in records, provided by s, viewed as temporal sequences locations and from which both popular attractions and visirs travel footprints can be extracted. Hence, through observing behaviors s from social media geo-tagged big data, popular attractions and many travelling laws could be effectively found, thus providing evidence and support for applications such as urism planning, urism resource development and intelligent travel recommendations. In recent years, researchers have proposed various approaches discover popular attractions from geo-tagged data, and spatial clustering algorithm is a common means among m [6,7]. There several reasons for using spatial clustering discover attractions. First all, we ten find it difficult get exact boundary data attractions, especially many attractions without walls have no exact boundaries, such as People s Heroes Monument and Eiffel Tower. Thus, through spatial clustering original user locations, generating natural boundaries attractions is a better choice from perspective human activities. Second, clustering-generated attractions, which naturally rich in human activity information (including users, time, space and visiting frequency), enable performing popularity analysis attractions and extracting travelers spatio-temporal patterns or travel laws. The potential extracted knowledge could provide decision support urism management department and play an important role in both urism planning and urism resource development. In addition, from individual view, travel preferences able be extracted from clustering-generated attractions, and thus, intelligent travel recommendation services could be developed for promote realization smart urism. 2. Related Work A classical spatial clustering method applied in attraction discovering is mean-shift algorithm. For example, Crandal and Kurashima used mean-shift algorithm cluster geographic coordinates and extracted popular attractions [7,8]. Yin et al. used mean-shift algorithm cluster geo-tagged phos and predict locations [9]. The advantage mean-shift algorithm is that re is no need specify number classes in advance, but only bandwidth search neighborhood is specified, so that mean-shift clustering algorithm can find several clustering centers without much prior knowledge [7,10]. In addition, Kennedy et al. used k-means algorithm classify geometric labels (i.e., latitude and longitude) pictures and n obtained information scenic spots [11]. Anor more popular spatial clustering method is DBSCAN (Density-Based Spatial Clustering Applications with Noise) algorithm, which has advantage requiring less field knowledge and enables finding irregularly-shaped clusters [12];

3 ISPRS Int. J. Geo-Inf. 2017, 6, Kisilevich et al. introduced an adaptive density-based clustering method named P-DBSCAN [13], which is designed on basis DBSCAN algorithm. Moreover, Yang used a spectral clustering method in attraction mining [14,15]; advantage this method is that number clusters can aumatically be adjusted. Zheng and Yuan studied how mine points interest and popular attractions using GPS trajecry data and proposed a hierarchical algorithm [16 18]. The algorithm first deals with GPS trajecry data and extracts user s stay points. Then, hierarchical clustering stay points is performed, and thus, a graph structure is created. In hierarchy, higher clustering level is, more stay points contained in clusters. Lastly, places and as can be sorted at different levels so as get points interest, attractions and popular as. There was also some work reveal how add semantic information attractions on basis spatial clustering. For example, Cao et al. studied how find attractions from phos with geographical location information and extracted representative phos and corresponding text labels, allowing users use pictures or text search attractions [19]. Gao extracted user knowledge from existing travel recommendation website (Yahoo Travel) and n enhanced text semantics phos [20]. Using knowledge that has been excavated from existing travel sites, it is possible measure wher text label pho is related travel and spread existing labels scenic spot phograph, thus increasing semantics attractions. Spatial clustering is core issue in finding popular attractions from geo-tagged data. In clustering algorithms described above, DBSCAN is widely used in spatial clustering as an excellent density-based clustering algorithm. Zhou used DBSCAN algorithm in his work predicting travel destination [21]. Lee adopted DBSCAN algorithm in work attraction popularity analysis [22]. Moreover, DBSCAN spatial clustering method, which is frequently used discover attractions, has become basis furr applications such as travel trajecry extraction, travel pattern analysis and intelligent attraction recommendation. For example, Cai extracted semantic trajecries from geo-tagged dataset after using DBSCAN as attraction-discovering method [22]. Vu and Chen performed analysis travel behaviors and movement patterns after finding attractions from geo-tagged phos [23,24]. Memon and Lee studied personalized travel recommendation approaches using Flickr geo-tagged dataset [25,26]. However, because geo-tagged social media data more complex in geographical distribution, clusters diverse in shape and ten contain many noisy data, which brings great challenges traditional spatial clustering algorithm. There many shortcomings as widely-used density-based clustering algorithm like DBSCAN. On one hand, re is a need determine fixed clustering thresholds (e.g., density and cluster number) in advance, which results in difficulty adapt clustering scenes when regional density has a large difference. On or hand, it is difficult distinguish adjacent high-density as, resulting in multiple different categories being assigned same cluster class. Therefore, this paper proposes a new popular attraction discovery approach based on Flickr geo-tagged phos, taking in account irregular shape clustering a and uneven distribution original coordinate points. Popular attractions extracted effectively through spatial clustering geo-location dataset and text mining methods. Then, popularity analysis extracted attractions is performed in different contexts. 3. Methodology 3.1. Data Pre-Processing We use Flickr public API obtain a tal 213,938 geo-tagged phos and related meta tag data from 1 January January 2016, in Beijing, China. The data span 11 years and come from 22,354 users worldwide. Then, Flickr dataset is preprocessed remove noisy data. Here, it mainly refers removing redundancy phos; a person may phograph a few

4 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, times in a short time at same place, and se phos upload on Flickr website in all. For situation, this situation, redundancy redundancy should be should removed. be removed. The phos The taken phos by taken same byuser within same user one hour within in one same hour in place should same place be treated shouldas beone. treated After asredundancy one. After redundancy processing, processing, number number pictures is pictures 185,531. is Figure 185, shows Figure 1 shows geographical geographical distribution distribution Flickr geo-tagged Flickr geo-tagged points after points removing after removing redundancy. redundancy. Figure 1. Flickr geo-coordinates distribution map. Figure 1. Flickr geo-coordinates distribution map. In addition, in order analyze heat distribution attractions in different time and wear In addition, conditions in(i.e., order contexts), analyze we also heat use distribution hisrical wear data attractions Beijing, in which different include time and temperature wear conditions and wear (i.e., condition contexts), information, we also use provided hisrical by wear Wunderground data Beijing, website which include ( temperature and wear condition information, provided by Wunderground website ( Spatial Clustering 3.2. Spatial Clustering new spatial clustering method is used achieve attraction discovering. First all, A new spatial clustering method is used achieve attraction discovering. First all, using using idea CFSFDP (Clustering by Fast Search and Find Density Peaks) proposed by idea CFSFDP (Clustering by Fast Search and Find Density Peaks) proposed by Rodriguez and Rodriguez and Laio [27], cluster centers extracted adaptively. Then, in order achieve initial Laio [27], cluster centers extracted adaptively. Then, in order achieve initial clustering, clustering, remaining points classified extracted clusters. The specific approach is: remaining points classified extracted clusters. The specific approach is: unclassified point i unclassified point i belongs category point whose density is greater than point i and belongs category point whose density is greater than point i and distance from point i distance from point i is shortest; after recursion, remaining points assigned extracted is shortest; after recursion, remaining points assigned extracted cluster centers. cluster centers. The selection cluster center is determined by two parameters point density ρ and The selection cluster center is determined by two parameters point density ρ relative distance δ (i.e., point that has large product value ρδ is chosen as cluster center). and relative distance δ (i.e., point that has large product value ρδ is chosen as cluster Equations (1) and (3) illustrate method calculating point density ρ and relative distance δ. center). Equations (1) and (3) illustrate method calculating point density ρ and relative distance δ. n ρ i = f ( d ij r ) (1) j=0 ρ = d { r (1) 1, x < 0 f (x) = (x) = 1, x < (2) 0, x 0 (2) 0, x ( 0 ) δ i = min j:ρj >ρ δ =min i dij (3) : (d ) (3) where ρ i represents density point i, δ i represents shortest distance between point i and where ρ represents density point i, δ represents shortest distance between point i and point whose density is greater than point i, d ij is distance from point i or point j in point whose density is greater than point i, d is distance from point i or point j in clustering region and r is search radius. clustering region and r is search radius. There two implied assumptions. First, cluster center has high-density, while being surrounded by low-density as, and second, cluster center has a relative large distance from or high-density centers. Since density ρ and relative distance δ taken in account

5 ISPRS Int. J. Geo-Inf. 2017, 6, There two implied assumptions. First, cluster center has high-density, while being surrounded ISPRS Int. J. Geo-Inf. 2017, 6, by low-density as, and second, cluster center has a relative large distance from or high-density centers. ger Since determine density cluster ρ andcenters, relative traditional distance δdbscan takenalgorithm s in accountshortcoming ger determine having cluster difficulty centers, distinguishing traditional adjacent DBSCAN high-density algorithm s as shortcoming can be overcome. having In addition, difficulty in distinguishing order solve adjacent problem high-density large differences as can in be overcome. density distribution In addition, in order clustering solve region, problem we use large road differences network divide in density classification distribution region in clustering several zones region, (see we Figure use 2), road and network density and divide relative classification distance region points in several standardized zones (see Figure in each 2), zone. and Hence density zoning and and relative standardization distance steps points enables our standardized approach in be each more zone. adaptive Hence zoning clustering and standardization scenes when steps regional enables density our approach has a large difference. be more adaptive Figure 2 illustrates clustering zoning scenesresult whenby main regional roads density in Beijing, haswhich a largehas difference. exactly Figure 144 zones. 2 illustrates zoning result by main roads in Beijing, which has exactly 144 zones. Figure 2. Zoning result by main roads in Beijing. Figure 2. Zoning result by main roads in Beijing Class Mergence 3.3. Class Mergence In above clustering process, threshold ρδ can be adjusted set discovering In above clustering process, threshold ρδ can be adjusted set discovering granularity attractions: lower threshold ρδ is set, more micro-scale clusters granularity attractions: lower threshold ρδ is set, more micro-scale clusters will be found. In this case, due some reasons, e.g., some places have large geographical a or will be found. In this case, due some reasons, e.g., some places have a large geographical a or individual behavior data mselves quite sparse, se may lead same place being individual behavior data mselves quite sparse, se may lead same place being partitioned in several different clusters. In order solve this problem, following steps partitioned in several different clusters. In order solve this problem, following steps performed achieve class mergence: performed achieve class mergence: (1) First, we use classical text similarity measure algorithm TF-IDF (Term Frequency-Inverse (1) First, we use classical text similarity measure algorithm TF-IDF (Term Frequency-Inverse Document Frequency) calculate semantic tag s weight for each initial cluster (in addition Document Frequency) calculate semantic tag s weight for each initial cluster (in addition geo-tagged information, re many semantic text tags in Flickr phos). Equation (4) geo-tagged information, re many semantic text tags in Flickr phos). Equation (4) illustrates weight calculation method. illustrates weight calculation method. h (t ) =tf(t ) idf(t ) =tf ( ) log(n/df(t )) (4) weight(t i ) = tf(t i ) idf(t i ) = tf j(ti ) log(n/df(t i )) (4) where tf ( ) represents frequency current tag t in cluster j, N is tal number clusters where and tfdf(t j(ti ) represents ) indicates how frequency many clusters have current current tag t i intag cluster t. j, N is tal number clusters and df(t i ) indicates how many clusters have current tag t i. (2) Then, multidimensional tag vecr is generated for each cluster, and cosine similarity (2) Then, a multidimensional tag vecr is generated for each cluster, and cosine similarity between adjacent clusters is calculated, reby merging adjacent clusters that actually have between adjacent clusters is calculated, reby merging adjacent clusters that actually have high similarity with each or. high similarity with each or. (X, Y) = cos = = x y (5) ( ) sim(x, Y) = cos θ = ( ) x y = i=1 n x i y i (5) where X and Y represent two clusters be compd, and i=1 n (x i) 2 ni=1 multidimensional (y i) 2 tag vecrs X and Y, respectively, and and indicate TF-IDF weights tag t in and, respectively. where X and Y represent two clusters be compd, x and y multidimensional tag vecrs X and Y, respectively, and x 3.4. Label Annotation for Popular Tourist Attractions i and y i indicate TF-IDF weights tag t i in x and y, respectively. The tags that have a p 10 TF-IDF weight value assigned as semantic labels each cluster. Then, name attribute cluster is determined by following method: a set Points Of Interest (POI) is obtained through calling Baidu s anti-geocoding API interface by inputting point coordinates in cluster, and name POI with highest number occurrences is assigned

6 ISPRS Int. J. Geo-Inf. 2017, 6, Label Annotation for Popular Tourist Attractions The tags that have a p 10 TF-IDF weight value assigned as semantic labels each cluster. Then, ISPRS Int. J. name Geo-Inf. attribute 2017, 6, 216 cluster is determined by following method: a set Points 6 16 Of Interest (POI) is obtained through calling Baidu s anti-geocoding API interface by inputting point cluster. Finally, if selected POI type is a attraction, cluster is retained, orwise coordinates in cluster, and name POI with highest number occurrences is assigned cluster is removed. As a result, attractions containing semantic information formed. cluster. Finally, if selected POI type is a attraction, cluster is retained, orwise Figure ISPRS 3 illustrates Int. J. Geo-Inf. two 2017, 6, word 216 cloud examples ( Forbidden City and Bird s 6 16 Nest), which cluster is removed. As a result, attractions containing semantic information formed. generated by user-annotated Flickr pho tags ger with corresponding TF-IDF Figure 3 illustrates cluster. Finally, two word if selected cloud examples POI type is a ( Forbidden attraction, City cluster and is retained, Bird s orwise Nest), which weights. The p 10 tags retained as label collection each cluster. Through labels, we generated by cluster user-annotated is removed. As a Flickr result, pho tags attractions ger containing with semantic corresponding information formed. TF-IDF weights. can understand Figure attitudes 3 illustrates and two interests word cloud examples visirs ( Forbidden attractions, City and which Bird s facilitates Nest), which building up The p 10 tags retained as label collection each cluster. Through labels, we can understand portrait generated by attraction user-annotated and Flickr providing pho tags decision ger support with corresponding urism TF-IDF management attitudesweights. and interests The p 10 tags visirs retained as attractions, label collection which each facilitates cluster. Through building labels, up we portrait department in both urism planning and urism resource development. attraction can understand providing attitudes and decision interests support visirs attractions, urismwhich management facilitates building department up in both portrait attraction and providing decision support urism management urism planning and urism resource development. department in both urism planning and urism resource development. Figure 3. Word cloud examples generated by user-annotated Flickr pho tags. Figure 3. Word cloud examples generated by user-annotated Flickr pho tags. Figure 3. Word cloud examples generated by user-annotated Flickr pho tags. 4. Popular Tourist Attraction-Discovering Result Popular Popular Tourist Tourist Attraction-Discovering Result 4.1. Result Description 4.1. Result Description Based on above methods, we use Java implement a attraction extraction ol. Flickr 4.1. Result Description geo-tagged pho data in Beijing from used as input. After spatial clustering Based on(step 1.2), above 300 clusters methods, generated, we use Java and n, implement through class mergence attraction step, extraction number ol. Flickr Based on clusters above is reduced methods, 243; finally, we use after Java semantic implement annotation, a number attraction clusters with extraction type ol. Flickr geo-tagged geo-tagged pho pho data data in attractions in Beijing Beijing from is decreased from used as input. After spatial clustering Figure illustrates used cluster as center input. distribution After map spatial after clustering (Step (Step 1.2), 1.2), clusters clusters attraction generated, discovering, generated, and and Figure and n, 5 n, shows through through heat map class class mergence attraction-visiting mergence step, step, activities. number number clusters is clusters reduced is reduced 243; finally, 243; after finally, semantic after semantic annotation, annotation, number number clusters with clusters with type type attractions attractions is decreased is decreased 143. Figure Figure illustrates 4 illustrates cluster center cluster distribution center distribution map after map after attraction attraction discovering, discovering, and Figure and 5Figure shows 5 shows heat map heat map attraction-visiting attraction-visiting activities. activities. Figure 4. Cluster center distribution map after attraction discovering. Figure 4. Cluster center distribution map after attraction discovering. Figure 4. Cluster center distribution map after attraction discovering.

7 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, Figure 5. Heat map attraction-visiting activities. Figure 5. Heat map attraction-visiting activities. Our attraction-discovering result illustrates that proposed approach is adaptive scenes Figure 5. Heat map attraction-visiting activities. Our attraction-discovering proposed approach is adaptive uneven density distribution result in illustrates clusteringthat region. The extracted attractions scenes more uneven density distribution in clustering region. The extracted attractions more evenly distributed in clustering region, wher in proposed high or low density as. This isscenes due Our attraction-discovering result illustrates that approach is adaptive evenly distributed in clustering region, wher in high or low density as. This is due zoning and standardization steps. More importantly, our approach can also effectively distinguish uneven density distribution in clustering region. The extracted attractions more zoning anddistributed standardization steps. importantly, can also effectively distinguish adjacent high-density as. TheMore original point coordinates Tiananmen a evenly in clustering region, wher in our highapproach or low density as. This is in due center adjacent high-density as. The original point coordinates Tiananmen a in center zoning steps. Morespatial importantly, our approach also effectively Beijing and quitestandardization intensive. Traditional clustering methodscan make it difficultdistinguish distinguish adjacent high-density as. The original pointclustering coordinates Tiananmen in enables good Beijing quite intensive. Traditional spatial methods make it a difficult center distinguish different attractions. However, Figure 6 indicates that our approach Beijing quite intensive. Traditional spatial clustering methods make it difficult distinguish different attractions. However, Figure 6 indicates that our approach enables good distinguishing distinguishing high-density as. In Tiananmen a, Tiananmen, Tiananmen Squ different as. attractions. that Tiananmen our approach enables good high-density InMonument However, Tiananmen a, 6indicates Tiananmen, and even and even People s Hero Figure effectively separated and extracted.squ Figure 7 shows distinguishing high-density as. In Tiananmen a, Tiananmen, Tiananmen Squ People s Monument anditsextracted. Figure 7 shows heat map heat maphero attractions in effectively Tiananmenseparated Squ and surrounding a. and even People s Hero Monument effectively separated and extracted. Figure 7 shows attractions in Tiananmen Squ and its surrounding a. heat map attractions in Tiananmen Squ and its surrounding a. Figure 6. Cluster distribution map in Tiananmen Squ and its surrounding a. Figure 6. Cluster distribution map in Tiananmen Squ and its surrounding a. Figure 6. Cluster distribution map in Tiananmen Squ and its surrounding a.

8 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, Figure 7. Heat map attractions in Tiananmen Squ and its surrounding a. Figure 7. Heat map attractions in Tiananmen Squ and its surrounding a. Figure 7. Heat map attractions in Tiananmen Squ and its surrounding a. In addition, we also find that proposed approach enables more effectively discovering In addition, we that enables more has afind large geographical aapproach and an irregular In classdiscovering mergence step, Inattraction addition, that we also also find that proposed proposed approach enablesshape. more effectively effectively discovering attraction that has a large geographical a and an irregular shape. In class mergence step, TF-DIF method is used merge adjacent semantic similar clusters achieve this goal. Figure 8 attraction that has a large geographical a and an irregular shape. In class mergence step, TF-DIF method is used merge adjacent semantic similar clusters achieve this goal. Figure shows results is used preliminary Wangfujing a. There two where88 TF-DIF method mergeclustering adjacent semantic similar clusters achieve thisclusters, goal. Figure shows results preliminary clustering Wangfujing a. There two clusters, where triangle represents cluster center and points in different clusters identified by different shows results preliminary clustering Wangfujing a. There two clusters, where triangle represents cluster center and points in different clusters identified by different colors. Then, through TF-IDF cosine similarity calculation, se two clusters eventually merged triangle represents cluster center and points in different clusters identified by different colors. Then, through in onethen, (see Figure 9).TF-IDF colors. through TF-IDF cosine cosine similarity similarity calculation, calculation, se se two two clusters clusters eventually eventually merged merged in one (see Figure 9). in one (see Figure 9). Figure 8. Preliminary clustering outputs for Wangfujing a. Figure 8. Preliminary clustering outputs for Wangfujing a. Figure 8. Preliminary clustering outputs for Wangfujing a.

9 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, Figure 9. Final clustering outputs after class mergence. Figure 9. Final clustering outputs after class mergence Result Comparison Figure 9. Final clustering outputs after class mergence Result Comparison For sake quantitative comparison, we also implement a DBSCAN-based program For 4.2. Result sake Comparison quantitative comparison, we also implement a DBSCAN-based program achieve achieve attraction discovering from Flickr geo-tagged dataset, using widely-used attraction machine learning For sake discovering package quantitative from Sckitlearn. comparison, Flickr geo-tagged The clustering we also parameters implement dataset, a using set DBSCAN-based widely-used as follows: program machine search radius learning achieve package Sckitlearn. attraction The discovering clustering from parameters Flickr geo-tagged set as follows: dataset, using search widely-used radius r = 15 m, r = 15 m, and density threshold minpts = 20. Then, final number generated attractions and machine density learning threshold package minpts Sckitlearn. = 20. Then, clustering final parameters number generated set as follows: attractions search radius is 100, is 100, and attraction distribution map is shown in Figure 10. and r = 15 m, and attraction density distribution threshold minpts map = is20. shown Then, in Figure final number 10. generated attractions is 100, and attraction distribution map is shown in Figure 10. Figure 10. DBSCAN-based cluster center distribution map after attraction discovering. Figure 10. DBSCAN-based cluster center distribution map after attraction discovering. DBSCAN is a typical density-based clustering algorithm, which enables discovering cluster DBSCAN consisting is a high-density typical density-based connectivity clustering a. From algorithm, figure, it which is observed enables that discovering algorithm can cluster also find clusters arbitrary shapes in a. From feature space figure, containing it is observed noisy data. that However, algorithm its consisting a high-density connectivity a. From figure, it is observed that algorithm can also can also shortcomings find clusters obvious, arbitrary as well. shapes First, in results feature space DBSCAN containing method noisy concentrated data. However, in its high-density region, while extracted attractions by our approach more evenly shortcomings obvious, as well. First, results DBSCAN method concentrated in high-density region, while extracted attractions by our approach more evenly

10 ISPRS Int. J. Geo-Inf. 2017, 6, find clusters arbitrary shapes in feature space containing noisy data. However, its shortcomings obvious, ISPRS Int. J. as Geo-Inf. well. 2017, First, 6, 216 results DBSCAN method concentrated in high-density region, while extracted attractions by our approach more evenly distributed in clustering distributed region, in wher clustering high region, or wher low density in high as or low (seedensity Figureas 4). Secondly, (see Figure DBSCAN 4). Secondly, cannot makedbscan an effective cannot distinction make an effective for distinction adjacent high-density for adjacent as high-density (such asas Tiananmen (such as Tiananmen Squ and its Squ and its surrounding a). Figure 11 demonstrates that Tiananmen, Tiananmen Squ and surrounding a). Figure 11 demonstrates that Tiananmen, Tiananmen Squ and People s Heroes People s Heroes Monument merged and classified as a category after running DBSCAN Monument merged and classified as a category after running DBSCAN clustering. clustering. Figure 11. DBSCAN-based cluster distribution in Tiananmen Squ and its surrounding a. Figure 11. DBSCAN-based cluster distribution in Tiananmen Squ and its surrounding a. In addition, we randomly select 10,000 points from Flickr dataset and n manually mark In addition, attraction we (i.e., randomly class) information select 10,000 by points overlaying from points Flickron dataset Baidu and n map. manually A tal mark 99 attraction categories, (i.e., class) including information 98 by attractions overlaying and none-attractions, points Baidu annotated map. Aon tal test 99points categories, (3568 points for non-attractions, and or 6432 points assigned or 98 attractions). including 98 attractions and none-attractions, annotated on test points (3568 points for Thus, we calculate Overall Accuracy (OA) and Average Accuracy (AA) two clustering non-attractions, and or 6432 points assigned or 98 attractions). Thus, we calculate results, which generated by Our Proposed Method (OPM) and DBSCAN, respectively. Then, Overall accuracy Accuracy evaluation (OA) is performed and Average comp Accuracy two (AA) methods. OA two and AA clustering calculated results, as shown which generated in Equations by Our(6) Proposed (8). Method (OPM) and DBSCAN, respectively. Then, accuracy evaluation is performed comp two methods. OA and AA OA = calculated as shown in Equations (6) (8). (6) OA = cn AA = i=1 TP a i (6) N (7) a = AA = cn i=1 a i (8) (7) cn where TP represents true positives class a i = i (i.e., TP i number points which classified (8) correctly in class i), cn represents class number N i (99 in this testing scenario), N represents where tal TP i point represents number (10,000 true in positives this testing class scenario), i (i.e., a is number accuracy rate points class which i and N is classified correctly point innumber class i), cn class represents i. class number (99 in this testing scenario), N represents tal point number Figure (10, illustrates in this testing results scenario), accuracy a evaluation. It is observed that our proposed i is accuracy rate class i and N i is point method is significantly higher than DBSCAN method in classification accuracy (87.6% vs. 69.8% number class i. in OA, and 82.7% vs. 51.3% in AA). We note that AA gap is greater than OA gap. Because Figure 12 illustrates results accuracy evaluation. It is observed that our proposed DBSCAN method cannot distinguish adjacent high-density as and cannot find attractions method in low-density is significantly as, higher parts than real DBSCAN attractions method lost in or classification misclassified. accuracy This reason (87.6% leads vs. 69.8% a in OA, and relatively 82.7% large vs. AA 51.3% gap in between AA). We note two approaches. that AA gap is greater than OA gap. Because DBSCAN method cannot distinguish adjacent high-density as and cannot find attractions in low-density as, parts real attractions lost or misclassified. This reason leads a relatively large AA gap between two approaches.

11 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, % 87.6% 82.7% 80.0% Accuracy 69.8% evaluation 14,454 Accuracy Accuracy 4,606 14,454 4, % 60.0% 40.0% 80.0% 20.0% 60.0% 4,337 4,337 3,870 3,870 2,863 2,863 2,059 2,059 Accuracy evaluation 1,791 1, % 1,424 1,609 1,424 1, % 1,370 1,370 1,236 1,236 1,138 1,138 1,106 1,106 1,104 1, % 82.7% 40.0% Overall accuracy Average accuracy 20.0% OPM DBSCAN 0.0% Overall accuracy Average accuracy Figure 12. Accuracy evaluation Our Proposed Method (OPM) and DBSCAN method. Figure 12. Accuracy evaluation Our Proposed OPM DBSCAN Method (OPM) and DBSCAN method. 5. Application Scenario: Popularity Analysis Tourist Attractions 5. Application Based Figure Scenario: on 12. Accuracy extracted Popularity evaluation clusters, Analysis Our popularity Proposed Tourist Method analysis Attractions (OPM) and DBSCAN attractions method. in Beijing is Based performed. on On extracted one hand, clusters, we extracted popularity p 20 popular analysis attractions; attractions or in hand, Beijing is we 5. Application also use time Scenario: and wear Popularity data Analysis build different Tourist contexts Attractions performed. On one hand, we extracted p 20 popular observe attractions; popularity on distribution or hand, Based on attractions extracted in Beijing clusters, from multi-dimensions. popularity analysis In this paper, we select attractions Forbidden in Beijing City, is we also use time and wear data build different contexts observe popularity distribution Bird s performed. Nest, On Jingshan one Park hand, and we extracted Badaling Great p Wall 20 popular as typical attractions; perform on or popularity hand, attractions in Beijing from multi-dimensions. In this paper, we select Forbidden City, analysis we also use under time different and wear contexts. data build different contexts observe popularity distribution Bird s Nest, Jingshan Park and Badaling Great Wall as typical attractions perform popularity attractions in Beijing from multi-dimensions. In this paper, we select Forbidden City, analysis 5.1. Bird s under Top Nest, 20 Popular different Jingshan Tourist contexts. Park Attractions and Badaling Great Wall as typical attractions perform popularity analysis Through under statistical different processing contexts. geo-tagged points within attraction clusters, 5.1. Top 20 Popular Tourist Attractions attractions ranking in p 20 shown in Figure 13. From Flickr perspective, popular Through attractions 5.1. Top 20 statistical Popular in Beijing Tourist processing include: Attractions Forbidden geo-tagged City, Summer points Palace, within Tiantan attraction Park, Tiananmen clusters, Squ, attractions People s Through ranking Heroes statistical inmonument, p processing 20Shichahai, shown Wangfujing, geo-tagged Figure 13. points Tiananmen From within Squ, Flickr attraction perspective, Jingshan clusters, Park, popular Sanlitun, attractions Mutianyu attractions in Beijing Great ranking Wall, include: Beihai p Forbidden 20 Park, Nanluoguxiang shown City, in Figure Summer Lane, 13. From Palace, Qianmen, Flickr Tiantan Lama perspective, Temple, Park, Tiananmen Beijing popular 798 Art Squ, People s District, attractions Heroes Drum in Monument, Beijing Tower, include: Bird's Shichahai, Nest, Forbidden National Wangfujing, City, Theatre Summer and Tiananmen Badaling Palace, Tiantan Great Squ, Wall. Park, Jingshan Tiananmen Park, Squ, Sanlitun, Mutianyu People s Great Heroes Wall, Monument, Beihai Park, Shichahai, Nanluoguxiang Wangfujing, Lane, Tiananmen Qianmen, Squ, Lama Jingshan Temple, Park, Beijing Sanlitun, 798 Art Mutianyu Great Wall, Beihai Park, Nanluoguxiang Lane, Qianmen, Lama Temple, Beijing 798 Art District, Drum Tower, Bird's Nest, National Heat Theatre distribution and Badaling Great Wall. District, Drum Tower, Bird's Nest, National Theatre and Badaling Great Wall Heat distribution , , % Figure 13. Top 20 popular attractions in Beijing. Figure 13. Top 20 popular attractions in Beijing. Figure 13. Top 20 popular attractions in Beijing.

12 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, Popularity Analysis under Different Temporal Contexts 5.2. Popularity Analysis under Different Temporal Contexts 5.2. Obviously, Popularity Analysis time has under an Different important Temporal impactcontexts travel choice. Observing people s visiting Obviously, time has an important impact on travel choice. Observing people s visiting behavior attractions from temporal perspective, some laws travelling could be found. behavior Obviously, time attractions has an from important temporal impact perspective, on travel some choice. laws Observing travelling people s could be visiting found. We perform attraction popularity analysis divided by quarters, months, slack/busy seasons (for each behavior We perform attraction attractions popularity from analysis temporal divided perspective, by quarters, some months, laws slack/busy travelling seasons could be (for found. each year, slack season is from 15 November 15 March, and busy season is from 16 March We year, perform slack attraction season is popularity from 15 November analysis divided 15 March, by quarters, and months, busy season slack/busy is from seasons 16 March (for each November) and weekend/working days. year, November) slack and season weekend/working is from 15 November days. 15 March, and busy season is from 16 March 14 November) Figure 14 represents quarterly heat distribution map four typical attractions (Forbidden Figure and 14 weekend/working represents quarterly days. heat distribution map four typical attractions City, Bird s Nest, Jingshan Park and Badaling Great Wall). From figure, it is observed that in (Forbidden Figure City, 14 represents Bird s Nest, Jingshan quarterly Park heat and Badaling distribution Great map Wall). From four figure, typical it is attractions observed (Forbidden second and third quarters, visiting heat four typical attractions is significantly higher that in second City, Bird s and third Nest, quarters, Jingshan Park visiting and Badaling heat Great four Wall). typical From attractions figure, is it significantly is observed than that higher in than first second and first fourth and and third quarters. fourth quarters, quarters. Figure 15 Figure visiting illustrates 15 heat illustrates monthly four monthly typical heat distribution attractions heat distribution map significantly map four attractions. higher four than attractions. From first and figure, From fourth it can figure, quarters. be seen it can Figure that be seen 15 four illustrates that four attractions monthly usher attractions heat distribution in ir usher visir in map ir peak roughly visir four peak inattractions. April roughly and August From in April each and figure, August year. it can each be year. seen that four attractions usher in ir visir peak roughly in April and August each year. Figure 14. Heat distribution map divided by quarters. Figure 14. Heat distribution map divided by quarters. Figure 14. Heat distribution map divided by quarters. Figure 15. Heat distribution map divided by months. Figure 15. Heat distribution map divided by months. Figure 15. Heat distribution map divided by months. Figure 16 represents heat distribution map four attractions divided by slack season and Figure busy 16 season. represents From heat figure, distribution it observed map that four in attractions busy season, divided by visiting slack heat season and four Figure busy 16attractions season. represents From is significantly heat figure, distribution it higher observed map than that slack four in season. attractions busy season, In divided busy visiting by season, slack heat it is season about and four three busy four attractions season. times From visiting significantly number figure, it higher is observed than attractions that slack in in season. busy slack In season, season. busy Figure visiting season, 17 it heat illustrates is about four three heat four distribution attractions times map isvisiting significantly number four higher attractions thandivided attractions slackby season. working In slack days season. busy and season, weekend Figure it 17 isdays. about illustrates From three four heat figure, times distribution it observed visiting map number that in four weekend attractions attractions day, divided invisiting by slack working heat season. days Figure and four weekend 17 illustrates attractions days. From heat is distribution slightly figure, higher it map is than observed four working that attractions day. weekend The divided reason day, bywhy working visiting gap days heat working and weekend days four days. and weekend From attractions days figure, is itslightly not is observed obvious higher that might than in be that weekend working Flickr day. day, users The mostly visiting reason come why heatfrom gap four outside working attractions days China, and and weekend is slightly as for foreign days higher is than not passengers, obvious working ir might travelling day. be that The reason Flickr behaviors users why mostly not gap limited come working by from working/weekend days outside and weekend China, day. days and as is not for obvious foreign might passengers, be thatir Flickr travelling users mostly behaviors come fromnot limited outside by China, working/weekend and as for foreign day. passengers, ir travelling behaviors not limited by working/weekend day.

13 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, Figure 16. Heat distribution map divided by slack/busy seasons. Figure 16. Heat distribution map divided by slack/busy seasons. Figure 16. Heat distribution map divided by slack/busy seasons. Figure 16. Heat distribution map divided by slack/busy seasons. Figure 17. Heat distribution map divided by working/weekend days. Figure 17. Heat distribution map divided by working/weekend days. Figure 17. Heat distribution map divided by working/weekend days Popularity Analysis under Different Wear Contexts 5.3. Popularity Analysis under Different Wear Contexts 5.3. Popularity In addition Analysis Figure time, under 17. Different Heat wear distribution Wear has an important map Contexts divided impact by working/weekend on choice days. travel. Observing people s In addition visiting behavior time, wear attractions has an important from impact wear on perspective, choice some travel. laws Observing travelling people s could 5.3. In addition Popularity be visiting found, Analysis time, behavior as well. under wear We Different perform attractions has Wear anattraction important Contexts from popularity impact wear on perspective, analysis choice from some travel. laws two Observing aspects travelling people s could temperature visiting be found, ranges behavior as and well. wear We perform conditions. attractions attraction from popularity wearanalysis perspective, from some two laws aspects travelling In addition time, wear has an important impact on choice travel. Observing couldtemperature Figure 18 ranges represents and wear heat conditions. people s be found, visiting as well. behavior We perform attraction distribution attractions popularity from map wear analysis four perspective, attractions from divided two some aspects laws by travelling temperature ranges ranges Figure (where 18 represents hot is above 30 heat degrees distribution Celsius, map warm is between four attractions 18 and 30 divided degrees by Celsius, temperature could andbe wear found, conditions. as well. We perform attraction popularity analysis from two aspects cool is ranges between (where 5 and 18 hot degrees is above Celsius 30 degrees and cold Celsius, below warm 5 degrees is between Celsius). 18 and From 30 degrees figure, Celsius, it is observed cool is temperature Figure 18 represents ranges and wear heat conditions. distribution map four attractions divided by temperature ranges between that (where visirs 5 and usually 18 degrees visit Celsius and attractions cold is below when 5 degrees temperature Celsius). From is moderate figure, (such it is as observed Figure hot 18 represents is above 30 degrees heat distribution Celsius, map warm is between four attractions 18 and 30 divided degrees by temperature Celsius, warm cool or is that cool), visirs and people usually visit particularly attractions reluctant when travel under temperature hot is wear moderate context. (such as Figure warm 19 or between ranges 5 and (where 18 degrees hot is above Celsius 30 degrees and cold Celsius, is below warm 5 degrees is between Celsius). 18 and From 30 degrees figure, Celsius, it iscool observed is cool), illustrates and people heat distribution particularly map reluctant four travel attractions under divided hot by wear wear context. conditions Figure (i.e., 19 that visirs between usually 5 and 18 visit degrees Celsius and attractions cold is below when 5 degrees temperature Celsius). From is moderate figure, (such it is observed as warm or illustrates sunny/cloudy heat days, distribution rainy days map and snowy four days). attractions From divided figure, by it wear is observed conditions that in (i.e., cool), that sunny/cloudy andvisirs peopleusually day, days, particularly visit rainy visiting days reluctant attractions heat and snowy travel when four days). under From temperature attractions hot figure, wear is moderate is significantly it context. (such is observed Figure as higher that 19 warm than in illustrates or heat cool), sunny/cloudy rainy/snowy distribution and people day, and map particularly people visiting four heat particularly attractions reluctant four reluctant divided travel under attractions travel by wear hot under is significantly conditions wear context. snowy wear (i.e., higher sunny/cloudy Figure 19 context. than illustrates heat distribution map four attractions divided by wear conditions (i.e., days, rainy/snowy daysday, and and snowy people days). particularly From reluctant figure, it is travel observed under that snowy in wear sunny/cloudy context. day, sunny/cloudy days, rainy days and snowy days). From figure, it is observed that in visiting heat four attractions is significantly higher than rainy/snowy day, sunny/cloudy day, visiting heat four attractions is significantly higher than and people rainy/snowy particularly day, and people reluctant particularly travel under reluctant snowy travel under wear context. snowy wear context. Figure 18. Heat distribution map divided by temperature. Figure 18. Heat distribution map divided by temperature. Figure 18. Heat distribution map divided by temperature. Figure 18. Heat distribution map divided by temperature.

14 ISPRS Int. J. Geo-Inf. 2017, 6, ISPRS Int. J. Geo-Inf. 2017, 6, Figure 19. Heat distribution map divided by wear conditions. Figure 19. Heat distribution map divided by wear conditions. 6. Conclusions 6. Conclusions Through spatial clustering original user locations, both natural boundaries and human Through activity spatial information clustering original attractions user locations, generated, both which natural facilitate boundaries performing and popularity human activity analysis information attractions and attractions extracting generated, travelers spatio-temporal which facilitate patterns performing popularity travel laws. analysis In this paper, we attractions propose a and new extracting method for discovering travelers popular spatio-temporal attractions, patterns or travel which laws. extracts In this hotspots paper, through we propose integrating a new spatial method clustering for discovering and text mining popular approaches. Attraction attractions, which discovery extracts experiments hotspots through performed integrating based spatial on clustering Flickr geotagged and text images mining within approaches. urban Attraction a discovery Beijing from experiments performed In addition, based based on on Flickr finding geotagged results images scenic within hotspots, this urban paper a Beijing analyzes from popularity distribution In addition, laws based Beijing s on finding attractions resultsunder scenic different hotspots, temporal thisand paper analyzes wear contexts. popularity distribution laws Beijing s attractions under different temporal and wear Our contexts. proposed attraction-discovering approach includes three major stages: spatial clustering, class mergence and label annotation. The innovations this approach include: (1) We first Our proposed attraction-discovering approach includes three major stages: spatial apply new clustering method CFSFDP in attraction discovering, and in process clustering, class mergence and label annotation. The innovations this approach include: (1) We first spatial clustering, zoning and standardization step is added make our approach more adaptive apply new clustering method CFSFDP in attraction discovering, and in process clustering scenes when regional density has large difference. Therefore, in clustering spatial clustering, zoning and standardization step is added make our approach more adaptive stage, our approach is completely different from traditional attraction-discovering methods, among clustering which DBSCAN scenes when is most regional popular density clustering has algorithm. large difference. (2) In addition Therefore, new in clustering clustering stage, method, our approach we also use is completely TF-IDF different method from generate traditional tag vecr attraction-discovering for each initial cluster methods, and n among which perform DBSCAN vecr issimilarity most popular calculation clustering merge algorithm. adjacent semantically (2) In addition similar clusters, new clustering which enables method, we more also use effective TF-IDF discovering method generate attraction tag vecr that has fora each large initial geographical cluster and a n and perform an irregular vecr similarity shape. This calculation is benefit merge adjacent class mergence semantically stage, and similar traditional clusters, attraction-discovering which enables more methods effective discovering never used combination attraction spatial that clustering has a largeand geographical class mergence. a and On an irregular whole, we shape. propose This a is unique benefitapproach class for mergence discovering stage, popular and traditional attractions, attraction-discovering which tally differs methods from never traditional used combination DBSCAN-based spatial discovery clustering method. and class The mergence. experimental On results whole, show we that propose compd a unique with approach fortraditional discovering DBSCAN popularmethod, attractions, this novel approach which tally is significantly differs from higher traditional in classification DBSCAN-based accuracy, discovery enables method. distinguishing The experimental adjacent high-density results show as and that has compd better adaptability with traditional in case DBSCAN an method, uneven this density novel distribution. approach is significantly higher in classification accuracy, enables distinguishing Moreover, through observing behaviors s from social media geo-tagged big data, adjacent high-density as and has better adaptability in case an uneven density distribution. many travelling laws can be effectively found, thus providing evidence and support for applications Moreover, through observing behaviors s from social media geo-tagged big data, such as urism planning and intelligent travel recommendations. In future, on one hand, we many travelling laws can be effectively found, thus providing evidence and support for applications hope explore optimization methods attraction discovering based on massive such as urism planning and intelligent travel recommendations. In future, on one hand, geo-tagged dataset; on or hand, urism recommendation research based on existing we attractions hope explore and laws is optimization expected be methods performed. attraction discovering based on massive geo-tagged dataset; on or hand, urism recommendation research based on existing attractions andacknowledgments: laws is expectedthis be research performed. was supported by grants from National Key Research and Development Program China (2017YFB ), National Natural Science Foundation China ( , , Acknowledgments: ), Scientific This research Research Key wasprogram supported Beijing by grants Municipal from Commission National Key Education Research (KM ) and Development Program and State China Key Laborary (2017YFB ), Resources and National Environmental NaturalInformation Science Foundation System. China ( , , ), Scientific Research Key Program Beijing Municipal Commission Education (KM ), NewAuthor Starting Contributions: Point ProgramHuang Beijing Z. conceived Union University and designed (ZK ), experiments. and State Key Peng Laborary X. performed Resources andexperiments Environmental and analyzed Information data. System. Peng X. and Huang Z. wrote paper.

15 ISPRS Int. J. Geo-Inf. 2017, 6, Author Contributions: Huang Z. conceived and designed experiments. Peng X. performed experiments and analyzed data. Peng X. and Huang Z. wrote paper. Conflicts Interest: The authors decl no conflict interest. References 1. Cao, L.; Luo, J.; Gallagher, A.; Jin, X.; Han, J.; Huang, T.S. A worldwide urism recommendation system based on geotagged web phos. In Proceedings IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP2010), Dallas, TX, USA, March 2010; pp Clements, M.; Serdyukov, P.; Vries, A.P.; Reinders, M.J. Using flickr geotags predict user travel Behavior. In Proceedings 33rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Geneva, Switzerland, July 2010; pp Ji, R.; Xie, X.; Yao, H.; Ma, W.Y. Mining city landmarks from blogs by graph modeling. In Proceedings 17th ACM International Conference on Multimedia, Beijing, China, Ocber 2009; pp Lu, X.; Wang, C.; Yang, J.M.; Pang, Y.; Zhang, L. Pho2trip: Generating travel routes from geo-tagged phos for trip planning. In Proceedings International Conference on Multimedia (MM2010), Firenze, Italy, Ocber 2010; ACM: New York, NY, USA, 2010; pp Wei, L.Y.; Zheng, Y.; Peng, W.C. Constructing popular routes from uncertain trajecries. In Proceedings 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD2012), Beijing, China, August 2012; ACM: New York, NY, USA, 2012; pp Comaniciu, D.; Meer, P. Mean shift: A robust approach ward feature space analysis. IEEE Trans. Pattern Anal. Mach. Intell. 2002, 24, [CrossRef] 7. Crandall, D.J.; Backstrom, L.; Huttenlocher, D.; Kleinberg, J. Mapping world s phos. In Proceedings 18th International Conference on World Wide Web, Madrid, Spain, April 2009; ACM: New York, NY, USA, 2009; pp Kurashima, T.; Iwata, T.; Me, G.; Fujimura, K. Travel route recommendation using geotags in pho sharing sites. In Proceedings 19th ACM International Conference on Information and Knowledge Management, Toron, ON, Canada, Ocber 2010; pp Yin, Z.; Cao, L. Diversified Trajecry Pattern Ranking in Geo-Tagged Social Media. In Proceedings SIAM International Conference on Data Mining, Mesa, AZ, USA, April Liu, J.; Zhou, T.; Wang, B. Research progress personalized recommendation system. Adv. Nat. Sci. 2009, 19, Kennedy, L.; Naaman, M.; Ahern, S.; Nair, R.; Rattenbury, T. How flickr helps us make sense world: context and content in community-contributed media collections. In Proceedings 15th International Conference on Multimedia, Berkeley, CA, USA, September Ye, M.; Yin, P.; Lee, W.C.; Lee, D.L. Exploiting geographical influence for collaborative point--interest recommendation. In Proceedings 34th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR2011), Beijing, China, July 2011; Association for Computing Machinery (ACM): New York, NY, USA, 2011; pp Kisilevich, S.; Mansmann, F.; Keim, D. P-DBSCAN: A density based clustering algorithm for exploration and analysis attractive as using collections geo-tagged phos. In Proceedings 1st International Conference and Exhibition on Computing for Geospatial Research & Application, Besda, MD, USA, June Yang, Y.; Gong, Z. Identifying points interest by self-tuning clustering. In Proceedings 34th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR2011), Beijing, China, July 2011; Association for Computing Machinery (ACM): New York, NY, USA, 2011; pp Shen, J.; Cheng, Z.; Shen, J.; Mei, T.; Gao, X. The evolution research on multimedia travel guide search and recommender systems. In MultiMedia Modeling; Springer: Berlin, Germany, 2014; pp Yuan, J.; Zheng, Y.; Xie, X. Discovering regions different functions in a city using human mobility and POIs. In Proceedings KDD-2012 Conference, Beijing, China, August 2012; pp

16 ISPRS Int. J. Geo-Inf. 2017, 6, Zheng, Y.; Zhang, L.; Xie, X.; Ma, W.Y. Mining interesting locations and travel sequences from GPS trajecries. In Proceedings International Conference on World Wide Web (WWW), Madrid, Spain, April 2009; ACM Press: New York, NY, USA, 2009; pp Zheng, V.W.; Zheng, Y.; Xie, X.; Yang, Q. Collaborative location and activity recommendations with GPS hisry data. In Proceedings WWW2010 Conference, Raleigh, NC, USA, April 2010; pp Gao, Y.; Tang, J.; Hong, R.; Dai, Q.; Chua, T.S.; Jain, R. W2Go: A travel guidance system by aumatic landmark ranking. In Proceedings International Conference on Multimedia, Firenze, Italy, Ocber 2010; pp Zhou, X.; Xu, C.; Kimmons, B. Detecting urism destinations using scalable geospatial analysis based on cloud computing platform. Comput. Environ. Urban Syst. 2015, 54, [CrossRef] 21. Lee, I.; Cai, G.; Lee, K. Mining Points--Interest Association Rules from Geo-tagged Phos. In Proceedings 46th Hawaii International Conference on System Sciences, Wailea, Maui, HI, USA, 7 10 January Cai, G.; Lee, K.; Lee, I. Discovering common semantic trajecries from geo-tagged social media. In Proceedings International Conference on Industrial, Engineering and Or Applications Applied Intelligent Systems, Morioka, Japan, 2 4 August 2016; Springer International Publishing: Cham, Switzerland, 2016; pp Vu, H.Q.; Li, G.; Law, R.; Ye, B.H. Exploring travel behaviors inbound s Hong Kong using geotagged phos. Tour. Manag. 2015, 46, [CrossRef] 24. Chen, S.; Yuan, X.; Wang, Z.; Guo, C.; Liang, J.; Wang, Z.; Zhang, X.L.; Zhang, J. Interactive visual discovering movement patterns from sparsely sampled geo-tagged social media data. IEEE Trans. Vis. Comput. Graph. 2016, 22, [CrossRef] [PubMed] 25. Memon, I.; Chen, L.; Majid, A.; Lv, M.; Hussain, I.; Chen, G. Travel recommendation using geo-tagged phos in social media for. Wirel. Pers. Commun. 2015, 80, [CrossRef] 26. Lee, I.; Cai, G.; Lee, K. Exploration geo-tagged phos through data mining approaches. Expert Syst. Appl. 2014, 41, [CrossRef] 27. Rodriguez, A.; Laio, A. Clustering by fast search and find density peaks. Science 2014, 344, [CrossRef] [PubMed] 2017 by authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under terms and conditions Creative Commons Attribution (CC BY) license (

A Framework of Detecting Burst Events from Micro-blogging Streams

A Framework of Detecting Burst Events from Micro-blogging Streams , pp.379-384 http://dx.doi.org/10.14257/astl.2013.29.78 A Framework of Detecting Burst Events from Micro-blogging Streams Kaifang Yang, Yongbo Yu, Lizhou Zheng, Peiquan Jin School of Computer Science and

More information

Exploring the Patterns of Human Mobility Using Heterogeneous Traffic Trajectory Data

Exploring the Patterns of Human Mobility Using Heterogeneous Traffic Trajectory Data Exploring the Patterns of Human Mobility Using Heterogeneous Traffic Trajectory Data Jinzhong Wang April 13, 2016 The UBD Group Mobile and Social Computing Laboratory School of Software, Dalian University

More information

Exploring Urban Areas of Interest. Yingjie Hu and Sathya Prasad

Exploring Urban Areas of Interest. Yingjie Hu and Sathya Prasad Exploring Urban Areas of Interest Yingjie Hu and Sathya Prasad What is Urban Areas of Interest (AOIs)? What is Urban Areas of Interest (AOIs)? Urban AOIs exist in people s minds and defined by people s

More information

GOVERNMENT GIS BUILDING BASED ON THE THEORY OF INFORMATION ARCHITECTURE

GOVERNMENT GIS BUILDING BASED ON THE THEORY OF INFORMATION ARCHITECTURE GOVERNMENT GIS BUILDING BASED ON THE THEORY OF INFORMATION ARCHITECTURE Abstract SHI Lihong 1 LI Haiyong 1,2 LIU Jiping 1 LI Bin 1 1 Chinese Academy Surveying and Mapping, Beijing, China, 100039 2 Liaoning

More information

Spatial Data Science. Soumya K Ghosh

Spatial Data Science. Soumya K Ghosh Workshop on Data Science and Machine Learning (DSML 17) ISI Kolkata, March 28-31, 2017 Spatial Data Science Soumya K Ghosh Professor Department of Computer Science and Engineering Indian Institute of Technology,

More information

Museumpark Revisit: A Data Mining Approach in the Context of Hong Kong. Keywords: Museumpark; Museum Demand; Spill-over Effects; Data Mining

Museumpark Revisit: A Data Mining Approach in the Context of Hong Kong. Keywords: Museumpark; Museum Demand; Spill-over Effects; Data Mining Chi Fung Lam The Chinese University of Hong Kong Jian Ming Luo City University of Macau Museumpark Revisit: A Data Mining Approach in the Context of Hong Kong It is important for tourism managers to understand

More information

Mining Frequent Trajectory Patterns and Regions-of-Interest from Flickr Photos

Mining Frequent Trajectory Patterns and Regions-of-Interest from Flickr Photos 2014 47th Hawaii International Conference on System Science Mining Frequent Trajectory Patterns and Regions-of-Interest from Flickr Photos Guochen Cai School of Business (IT) James Cook University Cairns

More information

Analysis of the Tourism Locations of Chinese Provinces and Autonomous Regions: An Analysis Based on Cities

Analysis of the Tourism Locations of Chinese Provinces and Autonomous Regions: An Analysis Based on Cities Chinese Journal of Urban and Environmental Studies Vol. 2, No. 1 (2014) 1450004 (9 pages) World Scientific Publishing Company DOI: 10.1142/S2345748114500043 Analysis of the Tourism Locations of Chinese

More information

Activity Identification from GPS Trajectories Using Spatial Temporal POIs Attractiveness

Activity Identification from GPS Trajectories Using Spatial Temporal POIs Attractiveness Activity Identification from GPS Trajectories Using Spatial Temporal POIs Attractiveness Lian Huang, Qingquan Li, Yang Yue State Key Laboratory of Information Engineering in Survey, Mapping and Remote

More information

A Geo-Statistical Approach for Crime hot spot Prediction

A Geo-Statistical Approach for Crime hot spot Prediction A Geo-Statistical Approach for Crime hot spot Prediction Sumanta Das 1 Malini Roy Choudhury 2 Abstract Crime hot spot prediction is a challenging task in present time. Effective models are needed which

More information

* Abstract. Keywords: Smart Card Data, Public Transportation, Land Use, Non-negative Matrix Factorization.

*  Abstract. Keywords: Smart Card Data, Public Transportation, Land Use, Non-negative Matrix Factorization. Analysis of Activity Trends Based on Smart Card Data of Public Transportation T. N. Maeda* 1, J. Mori 1, F. Toriumi 1, H. Ohashi 1 1 The University of Tokyo, 7-3-1 Hongo Bunkyo-ku, Tokyo, Japan *Email:

More information

Detecting Origin-Destination Mobility Flows From Geotagged Tweets in Greater Los Angeles Area

Detecting Origin-Destination Mobility Flows From Geotagged Tweets in Greater Los Angeles Area Detecting Origin-Destination Mobility Flows From Geotagged Tweets in Greater Los Angeles Area Song Gao 1, Jiue-An Yang 1,2, Bo Yan 1, Yingjie Hu 1, Krzysztof Janowicz 1, Grant McKenzie 1 1 STKO Lab, Department

More information

Clustering Analysis of London Police Foot Patrol Behaviour from Raw Trajectories

Clustering Analysis of London Police Foot Patrol Behaviour from Raw Trajectories Clustering Analysis of London Police Foot Patrol Behaviour from Raw Trajectories Jianan Shen 1, Tao Cheng 2 1 SpaceTimeLab for Big Data Analytics, Department of Civil, Environmental and Geomatic Engineering,

More information

Mobility Analytics through Social and Personal Data. Pierre Senellart

Mobility Analytics through Social and Personal Data. Pierre Senellart Mobility Analytics through Social and Personal Data Pierre Senellart Session: Big Data & Transport Business Convention on Big Data Université Paris-Saclay, 25 novembre 2015 Analyzing Transportation and

More information

Assessing pervasive user-generated content to describe tourist dynamics

Assessing pervasive user-generated content to describe tourist dynamics Assessing pervasive user-generated content to describe tourist dynamics Fabien Girardin, Josep Blat Universitat Pompeu Fabra, Barcelona, Spain {Fabien.Girardin, Josep.Blat}@upf.edu Abstract. In recent

More information

INDIANA ACADEMIC STANDARDS FOR SOCIAL STUDIES, WORLD GEOGRAPHY. PAGE(S) WHERE TAUGHT (If submission is not a book, cite appropriate location(s))

INDIANA ACADEMIC STANDARDS FOR SOCIAL STUDIES, WORLD GEOGRAPHY. PAGE(S) WHERE TAUGHT (If submission is not a book, cite appropriate location(s)) Prentice Hall: The Cultural Landscape, An Introduction to Human Geography 2002 Indiana Academic Standards for Social Studies, World Geography (Grades 9-12) STANDARD 1: THE WORLD IN SPATIAL TERMS Students

More information

Urban Population Migration Pattern Mining Based on Taxi Trajectories

Urban Population Migration Pattern Mining Based on Taxi Trajectories Urban Population Migration Pattern Mining Based on Taxi Trajectories ABSTRACT Bing Zhu Tsinghua University Beijing, China zhub.daisy@gmail.com Leonidas Guibas Stanford University Stanford, CA, U.S.A. guibas@cs.stanford.edu

More information

Geographical Bias on Social Media and Geo-Local Contents System with Mobile Devices

Geographical Bias on Social Media and Geo-Local Contents System with Mobile Devices 212 45th Hawaii International Conference on System Sciences Geographical Bias on Social Media and Geo-Local Contents System with Mobile Devices Kazunari Ishida Hiroshima Institute of Technology k.ishida.p7@it-hiroshima.ac.jp

More information

Urban Geo-Informatics John W Z Shi

Urban Geo-Informatics John W Z Shi Urban Geo-Informatics John W Z Shi Urban Geo-Informatics studies the regularity, structure, behavior and interaction of natural and artificial systems in the urban context, aiming at improving the living

More information

December 3, Dipartimento di Informatica, Università di Torino. Felicittà. Visualizing and Estimating Happiness in

December 3, Dipartimento di Informatica, Università di Torino. Felicittà. Visualizing and Estimating Happiness in : : Dipartimento di Informatica, Università di Torino December 3, 2013 : Outline : from social media Social media users generated contents represent the human behavior in everyday life, but... how to analyze

More information

Economic and Social Council 2 July 2015

Economic and Social Council 2 July 2015 ADVANCE UNEDITED VERSION UNITED NATIONS E/C.20/2015/11/Add.1 Economic and Social Council 2 July 2015 Committee of Experts on Global Geospatial Information Management Fifth session New York, 5-7 August

More information

Exploring spatial decay effect in mass media and social media: a case study of China

Exploring spatial decay effect in mass media and social media: a case study of China Exploring spatial decay effect in mass media and social media: a case study of China 1. Introduction Yihong Yuan Department of Geography, Texas State University, San Marcos, TX, USA, 78666. Tel: +1(512)-245-3208

More information

Exploiting Geographic Dependencies for Real Estate Appraisal

Exploiting Geographic Dependencies for Real Estate Appraisal Exploiting Geographic Dependencies for Real Estate Appraisal Yanjie Fu Joint work with Hui Xiong, Yu Zheng, Yong Ge, Zhihua Zhou, Zijun Yao Rutgers, the State University of New Jersey Microsoft Research

More information

World Geography. WG.1.1 Explain Earth s grid system and be able to locate places using degrees of latitude and longitude.

World Geography. WG.1.1 Explain Earth s grid system and be able to locate places using degrees of latitude and longitude. Standard 1: The World in Spatial Terms Students will use maps, globes, atlases, and grid-referenced technologies, such as remote sensing, Geographic Information Systems (GIS), and Global Positioning Systems

More information

Beating Social Pulse: Understanding Information Propagation via Online Social Tagging Systems 1

Beating Social Pulse: Understanding Information Propagation via Online Social Tagging Systems 1 Journal of Universal Computer Science, vol. 18, no. 8 (2012, 1022-1031 submitted: 16/9/11, accepted: 14/12/11, appeared: 28/4/12 J.UCS Beating Social Pulse: Understanding Information Propagation via Online

More information

DEVELOPMENT OF GPS PHOTOS DATABASE FOR LAND USE AND LAND COVER APPLICATIONS

DEVELOPMENT OF GPS PHOTOS DATABASE FOR LAND USE AND LAND COVER APPLICATIONS DEVELOPMENT OF GPS PHOTOS DATABASE FOR LAND USE AND LAND COVER APPLICATIONS An Ngoc VAN and Wataru TAKEUCHI Institute of Industrial Science University of Tokyo 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505 Japan

More information

Collaborative topic models: motivations cont

Collaborative topic models: motivations cont Collaborative topic models: motivations cont Two topics: machine learning social network analysis Two people: " boy Two articles: article A! girl article B Preferences: The boy likes A and B --- no problem.

More information

Revisitation in Urban Space vs. Online: A Comparison across POIs, Websites, and Smartphone Apps.

Revisitation in Urban Space vs. Online: A Comparison across POIs, Websites, and Smartphone Apps. 156 Revisitation in Urban Space vs. Online: A Comparison across POIs, Websites, and Smartphone Apps. HANCHENG CAO, ZHILONG CHEN, FENGLI XU, and YONG LI, Beijing National Research Center for Information

More information

Creating a Travel Brochure

Creating a Travel Brochure DISCOVERING THE WORLD! Creating a Travel Brochure Objective: Create a travel brochure to a well-known city including weather data and places to visit! Resources provided: www.weather.com, internet Your

More information

ST-DBSCAN: An Algorithm for Clustering Spatial-Temporal Data

ST-DBSCAN: An Algorithm for Clustering Spatial-Temporal Data ST-DBSCAN: An Algorithm for Clustering Spatial-Temporal Data Title Di Qin Carolina Department First Steering of Statistics Committee and Operations Research October 9, 2010 Introduction Clustering: the

More information

An Ontology-based Framework for Modeling Movement on a Smart Campus

An Ontology-based Framework for Modeling Movement on a Smart Campus An Ontology-based Framework for Modeling Movement on a Smart Campus Junchuan Fan 1, Kathleen Stewart 1 1 Department of Geographical and Sustainability Sciences, University of Iowa, Iowa City, IA, 52242,

More information

GIScience & Mobility. Prof. Dr. Martin Raubal. Institute of Cartography and Geoinformation SAGEO 2013 Brest, France

GIScience & Mobility. Prof. Dr. Martin Raubal. Institute of Cartography and Geoinformation SAGEO 2013 Brest, France GIScience & Mobility Prof. Dr. Martin Raubal Institute of Cartography and Geoinformation mraubal@ethz.ch SAGEO 2013 Brest, France 25.09.2013 1 www.woodsbagot.com 25.09.2013 2 GIScience & Mobility Modeling

More information

DM-Group Meeting. Subhodip Biswas 10/16/2014

DM-Group Meeting. Subhodip Biswas 10/16/2014 DM-Group Meeting Subhodip Biswas 10/16/2014 Papers to be discussed 1. Crowdsourcing Land Use Maps via Twitter Vanessa Frias-Martinez and Enrique Frias-Martinez in KDD 2014 2. Tracking Climate Change Opinions

More information

MACHINE LEARNING FOR MOBILITY STUDIES IN URBAN AND NATURAL AREAS TUULI TOIVONEN / DIGITAL GEOGRAPHY LAB

MACHINE LEARNING FOR MOBILITY STUDIES IN URBAN AND NATURAL AREAS TUULI TOIVONEN / DIGITAL GEOGRAPHY LAB MACHINE LEARNING FOR MOBILITY STUDIES IN URBAN AND NATURAL AREAS TUULI TOIVONEN / DIGITAL GEOGRAPHY LAB MACHINE LEARNING FOR MOBILITY STUDIES IN URBAN AND NATURAL AREAS TUULI TOIVONEN / DIGITAL GEOGRAPHY

More information

Inferring Friendship from Check-in Data of Location-Based Social Networks

Inferring Friendship from Check-in Data of Location-Based Social Networks Inferring Friendship from Check-in Data of Location-Based Social Networks Ran Cheng, Jun Pang, Yang Zhang Interdisciplinary Centre for Security, Reliability and Trust, University of Luxembourg, Luxembourg

More information

OPTIMAL ALLOCATION OF EMERGENCY SHELTER FACILITIES IN BEIJING. WU Wen-jie 1,2, ZHU Si-yuan 1,2 ZHANG Wen-zhong 1

OPTIMAL ALLOCATION OF EMERGENCY SHELTER FACILITIES IN BEIJING. WU Wen-jie 1,2, ZHU Si-yuan 1,2 ZHANG Wen-zhong 1 2010 4 114 :1003-2398(2010)04-0041-04 1,2 1,2 1,, (1., 100101;2., 100039) OPTIMAL ALLOCATION OF EMERGENCY SHELTER FACILITIES IN BEIJING WU Wen-jie 1,2, ZHU Si-yuan 1,2 ZHANG Wen-zhong 1 (1.Institute of

More information

GEOGRAPHIC INFORMATION SYSTEM AS A DECISION SUPPORT SYSTEM FOR TOURISM MANAGEMENT IN A DEVELOPING ECONOMY: A CASE OF ABUJA, NIGERIA

GEOGRAPHIC INFORMATION SYSTEM AS A DECISION SUPPORT SYSTEM FOR TOURISM MANAGEMENT IN A DEVELOPING ECONOMY: A CASE OF ABUJA, NIGERIA GEOGRAPHIC INFORMATION SYSTEM AS A DECISION SUPPORT SYSTEM FOR TOURISM MANAGEMENT IN A DEVELOPING ECONOMY: A CASE OF ABUJA, NIGERIA BY Surv. Dr. Kayode Odedare Department Geoinformatics, Federal School

More information

EXTRACTION OF REMOTE SENSING INFORMATION OF BANANA UNDER SUPPORT OF 3S TECHNOLOGY IN GUANGXI PROVINCE

EXTRACTION OF REMOTE SENSING INFORMATION OF BANANA UNDER SUPPORT OF 3S TECHNOLOGY IN GUANGXI PROVINCE EXTRACTION OF REMOTE SENSING INFORMATION OF BANANA UNDER SUPPORT OF 3S TECHNOLOGY IN GUANGXI PROVINCE Xin Yang 1,2,*, Han Sun 1, 2, Zongkun Tan 1, 2, Meihua Ding 1, 2 1 Remote Sensing Application and Test

More information

Introduction to Google Mapping Tools

Introduction to Google Mapping Tools Introduction to Google Mapping Tools Google s Mapping Tools Explore geographic data. Organize your own geographic data. Visualize complex data. Share your data with the world. Tell your story and educate

More information

Exploring the advances in using social media data for conservation science

Exploring the advances in using social media data for conservation science Exploring the advances in using social media data for conservation science Workshop at the 5th European Congress of Conservation Biology (ECCB), Jyväskylä, Finland Tuesday, June 12, 2018 11am - 1pm Room

More information

Temporal and Spatial Distribution of Tourism Climate Comfort in Isfahan Province

Temporal and Spatial Distribution of Tourism Climate Comfort in Isfahan Province 2011 2nd International Conference on Business, Economics and Tourism Management IPEDR vol.24 (2011) (2011) IACSIT Press, Singapore Temporal and Spatial Distribution of Tourism Climate Comfort in Isfahan

More information

Texas A&M University

Texas A&M University Texas A&M University CVEN 658 Civil Engineering Applications of GIS Hotspot Analysis of Highway Accident Spatial Pattern Based on Network Spatial Weights Instructor: Dr. Francisco Olivera Author: Zachry

More information

Road & Railway Network Density Dataset at 1 km over the Belt and Road and Surround Region

Road & Railway Network Density Dataset at 1 km over the Belt and Road and Surround Region Journal of Global Change Data & Discovery. 2017, 1(4): 402-407 DOI:10.3974/geodp.2017.04.03 www.geodoi.ac.cn 2017 GCdataPR Global Change Research Data Publishing & Repository Road & Railway Network Density

More information

A framework for spatio-temporal clustering from mobile phone data

A framework for spatio-temporal clustering from mobile phone data A framework for spatio-temporal clustering from mobile phone data Yihong Yuan a,b a Department of Geography, University of California, Santa Barbara, CA, 93106, USA yuan@geog.ucsb.edu Martin Raubal b b

More information

VISUALIZATION URBAN SPATIAL GROWTH OF DESERT CITIES FROM SATELLITE IMAGERY: A PRELIMINARY STUDY

VISUALIZATION URBAN SPATIAL GROWTH OF DESERT CITIES FROM SATELLITE IMAGERY: A PRELIMINARY STUDY CO-439 VISUALIZATION URBAN SPATIAL GROWTH OF DESERT CITIES FROM SATELLITE IMAGERY: A PRELIMINARY STUDY YANG X. Florida State University, TALLAHASSEE, FLORIDA, UNITED STATES ABSTRACT Desert cities, particularly

More information

The Challenge of Geospatial Big Data Analysis

The Challenge of Geospatial Big Data Analysis 288 POSTERS The Challenge of Geospatial Big Data Analysis Authors - Teerayut Horanont, University of Tokyo, Japan - Apichon Witayangkurn, University of Tokyo, Japan - Shibasaki Ryosuke, University of Tokyo,

More information

A Model of GIS Interoperability Based on JavaRMI

A Model of GIS Interoperability Based on JavaRMI A Model of GIS Interoperability Based on Java Gao Gang-yi 1 Chen Hai-bo 2 1 Zhejiang University of Finance & Economics, Hangzhou 310018, China 2 College of Computer Science and Technology, Zhejiang UniversityHangzhou

More information

Spatiotemporal Analysis of Urban Traffic Accidents: A Case Study of Tehran City, Iran

Spatiotemporal Analysis of Urban Traffic Accidents: A Case Study of Tehran City, Iran Spatiotemporal Analysis of Urban Traffic Accidents: A Case Study of Tehran City, Iran January 2018 Niloofar HAJI MIRZA AGHASI Spatiotemporal Analysis of Urban Traffic Accidents: A Case Study of Tehran

More information

Extracting Touristic Information from Online Image Collections

Extracting Touristic Information from Online Image Collections 2012 Eighth International Conference on Signal Image Technology and Internet Based Systems Extracting Touristic Information from Online Image Collections Edoardo Ardizzone *, Francesco Di Miceli **, Marco

More information

LABELING RESIDENTIAL COMMUNITY CHARACTERISTICS FROM COLLECTIVE ACTIVITY PATTERNS USING TAXI TRIP DATA

LABELING RESIDENTIAL COMMUNITY CHARACTERISTICS FROM COLLECTIVE ACTIVITY PATTERNS USING TAXI TRIP DATA LABELING RESIDENTIAL COMMUNITY CHARACTERISTICS FROM COLLECTIVE ACTIVITY PATTERNS USING TAXI TRIP DATA Yang Zhou 1, 3, *, Zhixiang Fang 2 1 Wuhan Land Use and Urban Spatial Planning Research Center, 55Sanyang

More information

IV Course Spring 14. Graduate Course. May 4th, Big Spatiotemporal Data Analytics & Visualization

IV Course Spring 14. Graduate Course. May 4th, Big Spatiotemporal Data Analytics & Visualization Spatiotemporal Data Visualization IV Course Spring 14 Graduate Course of UCAS May 4th, 2014 Outline What is spatiotemporal data? How to analyze spatiotemporal data? How to visualize spatiotemporal data?

More information

Mining Users Behaviors and Environments for Semantic Place Prediction

Mining Users Behaviors and Environments for Semantic Place Prediction Mining Users Behaviors and Environments for Semantic Place Prediction Chi-Min Huang Institute of Computer Science and Information Engineering National Cheng Kung University No.1, University Road, Tainan

More information

Diagnosing New York City s Noises with Ubiquitous Data

Diagnosing New York City s Noises with Ubiquitous Data Diagnosing New York City s Noises with Ubiquitous Data Dr. Yu Zheng yuzheng@microsoft.com Lead Researcher, Microsoft Research Chair Professor at Shanghai Jiao Tong University Background Many cities suffer

More information

A new metric of crime hotspots for Operational Policing

A new metric of crime hotspots for Operational Policing A new metric of crime hotspots for Operational Policing Monsuru Adepeju *1, Tao Cheng 1, John Shawe-Taylor 2, Kate Bowers 3 1 SpaceTimeLab for Big Data Analytics, Department of Civil, Environmental and

More information

Assessing the Influence of Spatio-Temporal Context for Next Place Prediction using Different Machine Learning Approaches

Assessing the Influence of Spatio-Temporal Context for Next Place Prediction using Different Machine Learning Approaches International Journal of Geo-Information Article Assessing the Influence of Spatio-Temporal Context for Next Place Prediction using Different Machine Learning Approaches Jorim Urner 1 ID, Dominik Bucher

More information

Expanding Typologies of Tourists Spatio-temporal Activities Using the Sequence Alignment Method

Expanding Typologies of Tourists Spatio-temporal Activities Using the Sequence Alignment Method Expanding Typologies of Tourists Spatio-temporal Activities Using the Sequence Alignment Method Junya Kawase a, and Fumiko Ito a a Department of Urban System Science Tokyo Metropolitan University, Japan

More information

Place this Photo on a Map: A Study of Explicit Disclosure of Location Information

Place this Photo on a Map: A Study of Explicit Disclosure of Location Information Place this Photo on a Map: A Study of Explicit Disclosure of Location Information Fabien Girardin 1, Josep Blat 1 1 Department of ICT, Pompeu Fabra University, 08003 Barcelona, Spain {Fabien.Girardin,

More information

A method of Area of Interest and Shooting Spot Detection using Geo-tagged Photographs

A method of Area of Interest and Shooting Spot Detection using Geo-tagged Photographs A method of Area of Interest and Shooting Spot Detection using Geo-tagged Photographs Motohiro Shirai Graduate School of Informatics, Shizuoka University 3-5-1 Johoku, Naka-ku Hamamatsu-shi, Shizuoka,

More information

International Journal of Computing and Business Research (IJCBR) ISSN (Online) : APPLICATION OF GIS IN HEALTHCARE MANAGEMENT

International Journal of Computing and Business Research (IJCBR) ISSN (Online) : APPLICATION OF GIS IN HEALTHCARE MANAGEMENT International Journal of Computing and Business Research (IJCBR) ISSN (Online) : 2229-6166 Volume 3 Issue 2 May 2012 APPLICATION OF GIS IN HEALTHCARE MANAGEMENT Dr. Ram Shukla, Faculty (Operations Area),

More information

Lesson 16: Technology Trends and Research

Lesson 16: Technology Trends and Research http://www.esri.com/library/whitepapers/pdfs/integrated-geoenabled-soa.pdf GEOG DL582 : GIS Data Management Lesson 16: Technology Trends and Research Overview Learning Objective Questions: 1. Why is integration

More information

Uncovering the Digital Divide and the Physical Divide in Senegal Using Mobile Phone Data

Uncovering the Digital Divide and the Physical Divide in Senegal Using Mobile Phone Data Uncovering the Digital Divide and the Physical Divide in Senegal Using Mobile Phone Data Song Gao, Bo Yan, Li Gong, Blake Regalia, Yiting Ju, Yingjie Hu STKO Lab, Department of Geography, University of

More information

Visitor Flows Model for Queensland a new approach

Visitor Flows Model for Queensland a new approach Visitor Flows Model for Queensland a new approach Jason. van Paassen 1, Mark. Olsen 2 1 Parsons Brinckerhoff Australia Pty Ltd, Brisbane, QLD, Australia 2 Tourism Queensland, Brisbane, QLD, Australia 1

More information

My Map Activity MINNESOTA SOCIAL STUDIES STANDARDS & BENCHMARKS

My Map Activity MINNESOTA SOCIAL STUDIES STANDARDS & BENCHMARKS My Map Activity OVERVIEW & OBJECTIVES Students will learn the basics of Google Maps while using geospatial data to create their neighborhood map with the places they spend time. They will also collect

More information

GIS Visualization: A Library s Pursuit Towards Creative and Innovative Research

GIS Visualization: A Library s Pursuit Towards Creative and Innovative Research GIS Visualization: A Library s Pursuit Towards Creative and Innovative Research Justin B. Sorensen J. Willard Marriott Library University of Utah justin.sorensen@utah.edu Abstract As emerging technologies

More information

ALTER ECO Alternative tourist strategies to enhance the local sustainable development of tourism by promoting Mediterranean identity Module 2: Testing

ALTER ECO Alternative tourist strategies to enhance the local sustainable development of tourism by promoting Mediterranean identity Module 2: Testing ALTER ECO Alternative tourist strategies to enhance the local sustainable development of tourism by promoting Mediterranean identity Module 2: Testing WHY? Areas of high tourism attraction in coastal cities

More information

Cell-based Model For GIS Generalization

Cell-based Model For GIS Generalization Cell-based Model For GIS Generalization Bo Li, Graeme G. Wilkinson & Souheil Khaddaj School of Computing & Information Systems Kingston University Penrhyn Road, Kingston upon Thames Surrey, KT1 2EE UK

More information

Large-scale Image Annotation by Efficient and Robust Kernel Metric Learning

Large-scale Image Annotation by Efficient and Robust Kernel Metric Learning Large-scale Image Annotation by Efficient and Robust Kernel Metric Learning Supplementary Material Zheyun Feng Rong Jin Anil Jain Department of Computer Science and Engineering, Michigan State University,

More information

Mapping the Urban Farming in Chinese Cities:

Mapping the Urban Farming in Chinese Cities: Submission to EDC Student of the Year Award 2015 A old female is cultivating the a public green space in the residential community(xiaoqu in Chinese characters) Source:baidu.com 2015. Mapping the Urban

More information

Cognitive Engineering for Geographic Information Science

Cognitive Engineering for Geographic Information Science Cognitive Engineering for Geographic Information Science Martin Raubal Department of Geography, UCSB raubal@geog.ucsb.edu 21 Jan 2009 ThinkSpatial, UCSB 1 GIScience Motivation systematic study of all aspects

More information

DATA SCIENCE SIMPLIFIED USING ARCGIS API FOR PYTHON

DATA SCIENCE SIMPLIFIED USING ARCGIS API FOR PYTHON DATA SCIENCE SIMPLIFIED USING ARCGIS API FOR PYTHON LEAD CONSULTANT, INFOSYS LIMITED SEZ Survey No. 41 (pt) 50 (pt), Singapore Township PO, Ghatkesar Mandal, Hyderabad, Telengana 500088 Word Limit of the

More information

Understanding Individual Daily Activity Space Based on Large Scale Mobile Phone Location Data

Understanding Individual Daily Activity Space Based on Large Scale Mobile Phone Location Data Understanding Individual Daily Activity Space Based on Large Scale Mobile Phone Location Data Yang Xu 1, Shih-Lung Shaw 1 2 *, Ling Yin 3, Ziliang Zhao 1 1 Department of Geography, University of Tennessee,

More information

VECTOR CELLULAR AUTOMATA BASED GEOGRAPHICAL ENTITY

VECTOR CELLULAR AUTOMATA BASED GEOGRAPHICAL ENTITY Geoinformatics 2004 Proc. 12th Int. Conf. on Geoinformatics Geospatial Information Research: Bridging the Pacific and Atlantic University of Gävle, Sweden, 7-9 June 2004 VECTOR CELLULAR AUTOMATA BASED

More information

1. Introduction. S.S. Patil 1, Sachidananda 1, U.B. Angadi 2, and D.K. Prabhuraj 3

1. Introduction. S.S. Patil 1, Sachidananda 1, U.B. Angadi 2, and D.K. Prabhuraj 3 Cloud Publications International Journal of Advanced Remote Sensing and GIS 2014, Volume 3, Issue 1, pp. 525-531, Article ID Tech-249 ISSN 2320-0243 Research Article Open Access Machine Learning Technique

More information

Key Issue 1: How Do Geographers Describe Where Things Are?

Key Issue 1: How Do Geographers Describe Where Things Are? Key Issue 1: How Do Geographers Describe Where Things Are? Pages 5-13 and some information from pages 15-18. ***Always keep your key term packet out whenever you take notes from Rubenstein. As the terms

More information

Spatial-Temporal Analytics with Students Data to recommend optimum regions to stay

Spatial-Temporal Analytics with Students Data to recommend optimum regions to stay Spatial-Temporal Analytics with Students Data to recommend optimum regions to stay By ARUN KUMAR BALASUBRAMANIAN (A0163264H) DEVI VIJAYAKUMAR (A0163403R) RAGHU ADITYA (A0163260N) SHARVINA PAWASKAR (A0163302W)

More information

Monitoring Urban Space Expansion Using Remote Sensing Data in Ha Long City, Quang Ninh Province in Vietnam

Monitoring Urban Space Expansion Using Remote Sensing Data in Ha Long City, Quang Ninh Province in Vietnam Monitoring Urban Space Expansion Using Remote Sensing Data in Ha Long City, Quang Ninh Province in Vietnam MY Vo Chi, LAN Pham Thi, SON Tong Si, Viet Key words: VSW index, urban expansion, supervised classification.

More information

The Changing Landscape of Land Administration

The Changing Landscape of Land Administration The Changing Landscape of Land Administration B r e n t J o n e s P E, PLS E s r i World s Largest Media Company No Journalists No Content Producers No Photographers World s Largest Hospitality Company

More information

An Automated Object-Oriented Satellite Image Classification Method Integrating the FAO Land Cover Classification System (LCCS).

An Automated Object-Oriented Satellite Image Classification Method Integrating the FAO Land Cover Classification System (LCCS). An Automated Object-Oriented Satellite Image Classification Method Integrating the FAO Land Cover Classification System (LCCS). Ruvimbo Gamanya Sibanda Prof. Dr. Philippe De Maeyer Prof. Dr. Morgan De

More information

City and SUMP of Ravenna

City and SUMP of Ravenna City and SUMP of Ravenna Nicola Scanferla Head of Mobility Planning Unit, Municipality of Ravenna nscanferla@comune.ra.it place your logo here 19 April, 2017 1st Steering Committee Meeting, Nicosia, Cyprus

More information

From Research Objects to Research Networks: Combining Spatial and Semantic Search

From Research Objects to Research Networks: Combining Spatial and Semantic Search From Research Objects to Research Networks: Combining Spatial and Semantic Search Sara Lafia 1 and Lisa Staehli 2 1 Department of Geography, UCSB, Santa Barbara, CA, USA 2 Institute of Cartography and

More information

Research Group Cartography

Research Group Cartography Research Group Cartography Research Group Cartography Towards supporting wayfinding LBS components 1. Mobile devices 2. Communication Network 3. Positioning Component 4. Service and Application Provider

More information

A Cloud Computing Workflow for Scalable Integration of Remote Sensing and Social Media Data in Urban Studies

A Cloud Computing Workflow for Scalable Integration of Remote Sensing and Social Media Data in Urban Studies A Cloud Computing Workflow for Scalable Integration of Remote Sensing and Social Media Data in Urban Studies Aiman Soliman1, Kiumars Soltani1, Junjun Yin1, Balaji Subramaniam2, Pierre Riteau2, Kate Keahey2,

More information

TEMPLATE FOR CMaP PROJECT

TEMPLATE FOR CMaP PROJECT TEMPLATE FOR CMaP PROJECT Project Title: Native Utah Plants Created by: Anna Davis Class: Box Elder 2008 Project Description Community Issue or Problem Selected -How project evolved? Community Partner(s)

More information

Belfairs Academy GEOGRAPHY Fundamentals Map

Belfairs Academy GEOGRAPHY Fundamentals Map YEAR 12 Fundamentals Unit 1 Contemporary Urban Places Urbanisation Urbanisation and its importance in human affairs. Global patterns of urbanisation since 1945. Urbanisation, suburbanisation, counter-urbanisation,

More information

Article An Improved Hybrid Method for Enhanced Road Feature Selection in Map Generalization

Article An Improved Hybrid Method for Enhanced Road Feature Selection in Map Generalization Article An Improved Hybrid Method for Enhanced Road Feature Selection in Map Generalization Jianchen Zhang 1,2,3, Yanhui Wang 1,2,3, *, and Wenji Zhao 1,2,3 1 College of Resources Environment and Tourism,

More information

Geo Business Gis In The Digital Organization

Geo Business Gis In The Digital Organization We have made it easy for you to find a PDF Ebooks without any digging. And by having access to our ebooks online or by storing it on your computer, you have convenient answers with geo business gis in

More information

ArcGIS GeoAnalytics Server: An Introduction. Sarah Ambrose and Ravi Narayanan

ArcGIS GeoAnalytics Server: An Introduction. Sarah Ambrose and Ravi Narayanan ArcGIS GeoAnalytics Server: An Introduction Sarah Ambrose and Ravi Narayanan Overview Introduction Demos Analysis Concepts using GeoAnalytics Server GeoAnalytics Data Sources GeoAnalytics Server Administration

More information

Twitter s Effectiveness on Blackout Detection during Hurricane Sandy

Twitter s Effectiveness on Blackout Detection during Hurricane Sandy Twitter s Effectiveness on Blackout Detection during Hurricane Sandy KJ Lee, Ju-young Shin & Reza Zadeh December, 03. Introduction Hurricane Sandy developed from the Caribbean stroke near Atlantic City,

More information

On the Problem of Error Propagation in Classifier Chains for Multi-Label Classification

On the Problem of Error Propagation in Classifier Chains for Multi-Label Classification On the Problem of Error Propagation in Classifier Chains for Multi-Label Classification Robin Senge, Juan José del Coz and Eyke Hüllermeier Draft version of a paper to appear in: L. Schmidt-Thieme and

More information

Models to carry out inference vs. Models to mimic (spatio-temporal) systems 5/5/15

Models to carry out inference vs. Models to mimic (spatio-temporal) systems 5/5/15 Models to carry out inference vs. Models to mimic (spatio-temporal) systems 5/5/15 Ring-Shaped Hotspot Detection: A Summary of Results, IEEE ICDM 2014 (w/ E. Eftelioglu et al.) Where is a crime source?

More information

Worksheet: The Climate in Numbers and Graphs

Worksheet: The Climate in Numbers and Graphs Worksheet: The Climate in Numbers and Graphs Purpose of this activity You will determine the climatic conditions of a city using a graphical tool called a climate chart. It represents the long-term climatic

More information

Geographic Knowledge Discovery Using Ground-Level Images and Videos

Geographic Knowledge Discovery Using Ground-Level Images and Videos This work was funded in part by a DOE Early Career award, an NSF CAREER award (#IIS- 1150115), and a PECASE award. We gratefully acknowledge NVIDIA for their donated hardware. Geographic Knowledge Discovery

More information

Where to Find My Next Passenger?

Where to Find My Next Passenger? Where to Find My Next Passenger? Jing Yuan 1 Yu Zheng 2 Liuhang Zhang 1 Guangzhong Sun 1 1 University of Science and Technology of China 2 Microsoft Research Asia September 19, 2011 Jing Yuan et al. (USTC,MSRA)

More information

Learning to Recommend Point-of-Interest with the Weighted Bayesian Personalized Ranking Method in LBSNs

Learning to Recommend Point-of-Interest with the Weighted Bayesian Personalized Ranking Method in LBSNs information Article Learning to Recommend Point-of-Interest with the Weighted Bayesian Personalized Ranking Method in LBSNs Lei Guo 1, *, Haoran Jiang 2, Xinhua Wang 3 and Fangai Liu 3 1 School of Management

More information

Spatial Information Retrieval

Spatial Information Retrieval Spatial Information Retrieval Wenwen LI 1, 2, Phil Yang 1, Bin Zhou 1, 3 [1] Joint Center for Intelligent Spatial Computing, and Earth System & GeoInformation Sciences College of Science, George Mason

More information

Oak Ridge Urban Dynamics Institute

Oak Ridge Urban Dynamics Institute Oak Ridge Urban Dynamics Institute Presented to ORNL NEED Workshop Budhendra Bhaduri, Director Corporate Research Fellow July 30, 2014 Oak Ridge, TN Our societal challenges and solutions are often local

More information

Abstract 1. The challenges facing Cartography in the new era.

Abstract 1. The challenges facing Cartography in the new era. Geo-Informatic Tupu the New development of Cartography Qin Jianxin Ph. D. The Research Center of Geographic Information System (GIS), Hunan Normal University. Changsha, Hunan, 410081 E-Mail: qjxzxd@sina.com

More information

Spatial Decision Tree: A Novel Approach to Land-Cover Classification

Spatial Decision Tree: A Novel Approach to Land-Cover Classification Spatial Decision Tree: A Novel Approach to Land-Cover Classification Zhe Jiang 1, Shashi Shekhar 1, Xun Zhou 1, Joseph Knight 2, Jennifer Corcoran 2 1 Department of Computer Science & Engineering 2 Department

More information

Encapsulating Urban Traffic Rhythms into Road Networks

Encapsulating Urban Traffic Rhythms into Road Networks Encapsulating Urban Traffic Rhythms into Road Networks Junjie Wang +, Dong Wei +, Kun He, Hang Gong, Pu Wang * School of Traffic and Transportation Engineering, Central South University, Changsha, Hunan,

More information

Discovery and Access of Geospatial Resources using the Geoportal Extension. Marten Hogeweg Geoportal Extension Product Manager

Discovery and Access of Geospatial Resources using the Geoportal Extension. Marten Hogeweg Geoportal Extension Product Manager Discovery and Access of Geospatial Resources using the Geoportal Extension Marten Hogeweg Geoportal Extension Product Manager DISCOVERY AND ACCESS USING THE GEOPORTAL EXTENSION Geospatial Data Is Very

More information