TRAFFIC RISK MINING. Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi

Transcription

1 Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi B.E. Computer Engineering, Karmaveer Kakasaheb Wagh Institute of Engineering, Education and Research, Nashik. ABSTRACT In recent years, a large amount of traffic-related data is obtained from various sources like manually and through sensors and social media which can be in the form of traffic statistics, accident statistics, road information, and users comments. Traffic risk refers to the possibility of occurrence of traffic accidents on different types of roads. Specifically, we focus on two issues: 1) predicting the number of accidents on any road or at intersection, 2) clustering roads to calculate risk factors for risky road clusters. Using traffic data in India, we will demonstrate that the proposed system can be used to predict traffic risk at any location more efficiently and accurately than existing methods, and that a number of clusters of risky roads can be identified and characterized by risk factor. We will be using a system which uses algorithms like K-means, KNN, feature-based matrix factorization and non-negative matrix factorization algorithms. Also, as the risk factor of all the location will be found, it can be arranged according to their values. The risk factor of clusters can be used to sort the locations. Keywords - Learning systems unsupervised learning, machine intelligence pattern analysis. [1] INTRODUCTION The data collected related to traffic is becoming increasingly varied and heterogeneous. Indeed, the data may also include information collected through a variety of sensors and social networks and not only traffic statistics. In recent years efforts are made to use such traffic data for a wider range of purposes, such as safety management which can be used by government, driver support, traffic infrastructure design, and disaster prevention. For example, a service can be designed to show various traffic statistics, such as the frequencies of accidents and braking at different locations on the road map. It can also show comments posted online by drivers and pedestrians for these locations. This information can be used by governments to identify high-risk locations and adopt safety measures accordingly. There are most importantly two problems related to information collected, they are as follows: 1) Need for Completing Traffic Data: The first problem is the unavailability of data for all of the locations on the map. Hence, it is difficult to calculate risk factor for that particular location regardless of their high potential risk. Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi 1

2 2) Need of Discovering Global Knowledge of Traffic Risk: The second problem is the isolated nature of the traffic risk information associated with each location, which complicates the development of any system for traffic risk management. If the traffic risk information could be shared among locations with similar road conditions, it would be possible to extract their common traffic risk patterns. Thus, it would be possible to understand risk involved in the roads with common features. Therefore, it is necessary to combine the information from all the locations in order to obtain a global perspective that would be useful for ranking the risk associated with different locations. [2] RELATED WORK Koichi Moriya, Shin Matsushima, and Kenji Yamanishi in Traffic Risk Mining From Heterogeneous Road Statistics used real-traffic data in Tokyo, and demonstrated that Feature-based non-negative matrix factorization algorithm can be used to predict traffic risk at any location more accurately and efficiently than existing methods, and that a number of clusters of risky roads can be identified and characterized by two risk factors. [1] E. Bayam et al. in Older drivers and accidents: A meta-analysis and data mining application on traffic accident data studied teenage driving and associated accidents thoroughly. This paper addresses these two needs by providing a meta-analysis of the existing literature on senior drivers and showing how data mining techniques could be used in this application. [2] T. Beshah and S. Hill in Mining road traffic accident data to improve safety: Role of roadrelated factors on accident severity in Ethiopia applied data mining technologies to link recorded road characteristics to accident severity in Ethiopia, and developed a set of rules that could be used by the Ethiopian Traffic Agency to improve safety. [3] L. Chang and W. Chen in Data mining of tree-based models to analyse freeway accident frequency explained that Classification and Regression Tree (CART) is one of the most widely applied data mining techniques, has been commonly employed in business administration, industry, and engineering. CART does not require any pre-defined underlying relationship between target (dependent) variable and predictors (independent variables) and has been shown to be a powerful tool, particularly for dealing with prediction and classification problems. [4] M. Chong, A. Abraham, and M. Paprzycki in Traffic accident analysis using machine learning paradigms summarizes the performance of four machine learning paradigms applied to modelling the severity of injury that occurred during traffic accidents. They considered neural networks trained using hybrid learning approaches, support vector machines, decision trees and a concurrent hybrid model involving decision trees and neural networks. Experiment results reveal that among the machine learning paradigms considered the hybrid decision tree-neural network approach outperformed the individual approaches. [6] I. S. Dhillon and S. Sra in Generalized non negative matrix approximations with Bregman divergences makes algorithmic progress by modelling and solving (using multiplicative updates) new generalized NNMA problems that minimize Bregman divergences between the input matrix and its low rank approximation. In addition, this paper shows how to use penalty functions for incorporating constraints other than non-negativity into the problem. Further, some interesting extensions to the use of link functions for modelling nonlinear relationships are also discussed. [7] Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi 2

3 J. Kim and H. Park in Sparse non negative matrix factorization for clustering studied the properties of Non negative Matrix Factorization (NMF) as a clustering method by relating its formulation to other methods such as K-means clustering. [11] S. Krishnaveni and M. Hemalantha in A perspective analysis of traffic accident using data mining techniques deals with the some of classification models to predict the severity of injury that occurred during traffic accidents. They compared Naive Bayes Bayesian classifier, AdaBoostM1 Meta classifier, PART Rule classifier, J48 Decision Tree classifier and Random Forest Tree classifier for classifying the type of injury severity of various traffic accidents. The final result showed that the Random Forest outperforms than other four algorithms. [12] D. D. Lee and H. S. Seung in Algorithms for non-negative matrix factorization discussed two different multiplicative algorithms for NMF are analysed. They differ only slightly in the multiplicative factor used in the update rules. One algorithm was shown to minimize the conventional least squares error while the other minimizes the generalized Kullback-Leibler divergence. The monotonic convergence of both algorithms was proven using an auxiliary function analogous to that used for proving convergence of the Expectation- Maximization algorithm. [13] D. D. Lee and H. S. Seung in Learning the parts of objects by non-negative matrix factorization demonstrated an algorithm for non-negative matrix factorization that is able to learn parts of faces and semantic features of text. This is in contrast to other methods, such as principal components analysis and vector quantization, that learn holistic, not parts-based, representations. [14] X. Li, J. Han, J. Lee, and H. Gonzalez in Traffic density-based discovery of hot routes in road networks tried finding hot routes (traffic flow patterns) in a road network is an important problem. They are beneficial to city planners, police departments, real estate developers, and many others. Knowing the hot routes allows the city to better direct traffic or analyse congestion causes. In the past, this problem has largely been addressed with domain knowledge of city. But in recent years, detailed information about vehicles in the road network have become available. With the development and adoption of RFID and other location sensors, an enormous amount of moving object trajectories are being collected and can be used towards finding hot routes. [15] M. Nakano, J. Le Roux, H. Kameoka, T. Nakamura, N.Ono, and S. Sagayama in Bayesian non parametric spectrogram modelling based on infinite factorial infinite hidden Markov model presents a Bayesian non parametric latent source discovery method for music signal analysis. In audio signal analysis, an important goal is to decompose music signals into individual notes, with applications such as music transcription, source separation or notelevel manipulation. Recently, the use of latent variable decompositions, especially nonnegative matrix factorization (NMF), has been a very active area of research. These methods are facing two, mutually dependent, problems: first, instrument sounds often exhibit timevarying spectra, and grasping this time-varying nature is an important factor to characterize the diversity of each instrument; moreover, in many cases we do not know in advance the number of sources and which instruments are played. Conventional decompositions generally fail to cope with these issues as they suffer from the difficulties of automatically determining the number of sources and automatically grouping spectra into single events. They addressed Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi 3

4 both these problems by developing a Bayesian non parametric fusion of NMF and hidden Markov model (HMM). Their model decomposes music spectrograms in an automatically estimated number of components, each of which consisting in an HMM whose number of states is also automatically estimated from the data. [17] [3] SUMMARY This paper gives us an idea of how a system can be designed for traffic risk mining. This system can use the method which is based on an algorithm that uses multiplicative updates of a variant of FMF, which is also considered an extension of NMF. In the system we can have three main objectives 1)it can be used for predicting number of accidents on a particular road 2)it can be used for clustering using the values of risk factor 3)using risk factor values clusters can be arranged or sorted. We also established methods of evaluation and characterization of the obtained clusters. REFERENCES [1] Koichi Moriya, Shin Matsushima, and Kenji Yamanishi, Traffic Risk Mining From Heterogeneous Road Statistics in Proceedings of International Conference on Data Science and Advanced Analytics (DSAA), Paris, France, December 2015, IEEE, Computer Society. [2] E. Bayam, J. Liebowitz, and W. Agresti, Older drivers and accidents: A meta-analysis and data mining application on traffic accident data, Expert Syst. Appl., vol. 29, no. 3, pp , [3] T. Beshah and S. Hill, Mining road traffic accident data to improve safety: Role of road-related factors on accident severity in Ethiopia, in Proc. AAAI Artif. Intell. Develop. (AI-D), [4] L. Chang and W. Chen, Data mining of tree-based models to analyse freeway accident frequency, J. Safety Res., vol. 36, no.4, pp , [5] T. Chen, Z. Zheng, Q. Lu, W. Zhang, and Y. Yu. (2011). Feature-based matrix factorization. [Online]. Available: [6] M. Chong, A. Abraham, and M. Paprzycki, Traffic accident analysis using machine learning paradigms, Informatica, vol. 29, no. 1,pp. 8998, [7] I. S. Dhillon and S. Sra, Generalized non negative matrix approximations with Bregman divergences, in Proc. Adv. NIPS, 2005, pp [8] J. Han, J.-G. Lee, H. Gonzalez, and X. Li, Mining massive RFID, trajectory, and traffic data sets, in Proc. IEEE ICDM Contest, TomTom Traffic Prediction Intell. GPS Navigat., 2008, p.2. [9] S. Hirai and K. Yamanishi, Efficient computation of normalized maximum likelihood codes for Gaussian mixture models with its application to clustering, in Proc. IEEE Int. Symp. Inf. Theory (ISIT), Jul.2011, pp [10] K. Ishiguro and K. Takeuchi, Extracting essential structure from data,ntt Tech. Rev., vol. 10, no. 11, pp. 16, Nov [11] J. Kim and H. Park, Sparse nonnegative matrix factorization for clustering, Georgia Inst. Technol., Atlanta, GA, USA,Tech.Rep. GT-CSE , Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi 4

5 [12] S. Krishnaveni and M. Hemalantha, A perspective analysis of traffic accident using data mining techniques, Int. J. Comput. Appl., vol. 23, no. 7, pp. 4048, Jun [13] D. D. Lee and H. S. Seung, Learning the parts of objects by non-negative matrix factorization, Nature, vol. 401, no. 6755, pp , [14] D. D. Lee and H. S. Seung, Algorithms for non-negative matrix factorization, in Proc. Adv. NIPS, 2001, pp [15] X. Li, J. Han, J. Lee, and H. Gonzalez, Traffic density-based discovery of hot routes in road networks, in Advances in Spatial and Temporal Databases. SSTD (Lecture Notes in Computer Science), vol Springer, 2007, pp [16] K. Moriya, S. Matsushima, and K. Yamanishi, Traffic risk mining from heterogeneous road statistics, in Proc. IEEE Int.Conf. Data Sci. Adv.Anal., Oct. 2015, pp [17] M. Nakano, J. Le Roux, H. Kameoka, T. Nakamura, N. Ono, and S. Sagayama, Bayesian nonparametric spectrogram modeling based on infinite factorial infinite hidden Markov model, in Proc. WASPAA, 2011,pp [18] J. Shang, Y. Zheng, W. Tong, E. Chang, and Y. Yu, Inferring gas consumption and pollution emission of vehicles throughout a city, in Proc. KDD, 2014, pp [19] H. Shinnou and M. Sasaki, Refinement of document clustering by using NMF, in Proc. PACLIC, 2007, pp [20] P. Smaragdis and J. C. Brown, Non-negative matrix factorization for polyphonic music transcription, in Proc. WASPAA, 2003, pp [21] W. Xu, X. Liu, and Y. Gong, Document clustering based on non-negative matrix factorization, in Proc. SIGIR, 2003, pp [22] K. Yamanishi, A learning criterion for stochastic rules, Mach. Learn.,vol.9, no. 2, pp , [23] S. Zhang, W. Wang, J. Ford, and F. Makedon, Learning from incomplete ratings using nonnegative matrix factorization, in Proc. 6th SIAM Conf. Data Mining (SDM), 2006, pp [24] W. J. Lawton and E. A. Sylvestre, Self modelling curve resolution, Technometrics, vol. 13, no. 3, pp , Gargee Chaudhari, Asawari Deore, Shruti Sonaje, Rabia Qazi 5