Agent-based distributed time series forecasting system

Size: px
Start display at page:

Download "Agent-based distributed time series forecasting system"

Transcription

1 Journal of Theoretical and Applied Computer Science Vol. 9, No. 1, 2015, pp ISSN (printed), (online) Agent-based distributed time series forecasting system Michał Zabłocki 1 1 Faculty of Computer Science and Information Technology, West Pomeranian University of Technology, Szczecin, Poland mzablocki@wi.zut.edu.pl Abstract: Many studies have demonstrated that agent-based distributed computing improves quality of distributed computations. In this paper, self-aware software agents are used to manage the distributed computations in order to improve effectiveness of investment decisions. A distributed time series forecasting approach based on the modified Group Method Data Handling (GMDH) method and agent oriented programing is proposed. The forecasted results computed by agents are used to make an investment decision. To assess the effectiveness of the system, we used the time series of EUR/USD currency pair stock prices. The empirical results with a real data set clearly suggest that the system can be deployed on the trading platform to automate process of the prediction of financial markets. Keywords: time series prediction, GMDH, stock market forecasting, multiagent system, distributed system 1. Introduction The current theory of economic forecasting is based almost only on the theory of probability and mathematical statistics. However, practice shows that this is not enough. Nowadays, it is believed that only use of sophisticated methods of modern technical analysis (using artificial neural networks, pattern recognition, genetic algorithms, etc.) makes sense during creation of complex models in order to reproduce the economic reality. Many Nobel Prize winners in Economics of last few years, despite the deliberately made mistakes (at simplifying the assumptions necessary to build the model), contributed to the significant knowledge increase about mathematical modeling. Wiliński [1] used GMDH (Group Method of Data Handling) [2] as an effective methods of forecasting economic time series. A basis for GMDH by the first time was proposed in [3][4][5][6][7] by Ivakhnenko. The method, in the part related to the forecasting model, is based on two principles: to build the best regression model, and to reduce the complexity of the regression model to the lowest level acceptable by a researcher. Regression model should take into account only a few of many arguments (input variables or other variables). This means that in many cases, it is necessary to examine even thousands of regression models. Every such model will have a different configuration of input arguments. The models that have the potential to give best forecast in the future are selected to be a part of more complex models. Therefore, the method can be considered as a self-organizing system. At the every step, the algorithm selects several (sometimes many)

2 18 Michał Zabłocki best models, instead of the best single model, in order to leave a freedom in decisionmaking at next step. Combinatorial Algorithm is the basic GMDH algorithm. In its original idea it was proposed by prof. Ivakhnenko. In the most general form of the algorithm is composed of multiple layers of active neurons, which resembles a neural network. The purpose of each layer is a selection of outputs from the previous layer and building a model based on external criteria. At each stage (the output of each layer), such model can be used for forecasting. Neuron activity is based on the selection of the internal standard pool of inputs (from the previous layer) to build its model meets the external criterion and can be used as input for the next layer. A set of models obtained by the use of GMDH can be broadly defined as a subset of Gabor-Kolmogorov polynomial. M y = α 0 + i=1 α i x i + i=1 M M j=1 M M M α ij x i x j + α ijk x i x j x k +... (1) i=1 j=1k=1 where: α 0 constant term of a polynomial, Α(α i, α ij, α ijk,...) vector of coefficients, Χ(x i, x j, x k,...) vector of variables. Combinatorial algorithm is relatively simple linear variation of the method, but in many cases very effective. The simplicity of the algorithm is seeming, as with the growth of variable numbers, the computational complexity increases rapidly and it is necessary to impose restrictions that result from the limited time and computational power. For this reason, it becomes necessary to create subsets of models, instead of searching all possible solutions. The GMDH, due to the fact that it requires several complex tasks to be performed (the construction, evaluation and selection of many regression models), in many cases requires huge computational power. This demand, in spite of the enormous technological progress in the field of computer technology, may not always be met by a single computer. In many cases, the forecasts that we want to build have a small horizon. So, for obvious reasons, the calculation time can not transcend that horizon. The solution to this problem may be scattering calculations on several computers. The proposed time series forecasting system is based on distributed agent-based solution. Namely, a few software agents cooperate with each other in order to compute the forecast. Such multiagent system gains new capabilities that gives agent-oriented programming (AOP) paradigm [8]. 2. Related works Nowadays, the GMDH algorithms are widely used and many modifications of them are still developed by many researchers. Below are presented quite new and most interesting applications of this model computation technique. In [9] authors compare the accuracy of GMDH Analogues Complexing as typical nonparametric method and the Group of Adaptive Models Evolution (GAME, also based on GMDH theory) as a parametric method. They use medical data from Motol hospital in Prague and horticulture data from Hort Research New Zealand. The results of their experiments showed that both methods have good performance. This paper [10] proposes a novel method of constructing forecasting models (GLSSVM), which combines the group method of data handling (GMDH) and the least squares support vector machine (LSSVM). The GMDH method is used to determine the useful input variables for LSSVM model which works as prediction method. The authors

3 Agent-based distributed time series forecasting system 19 study the application of GLSSVM for monthly river flow forecasting of Selangor and Barman River. The results point that the proposed solution is a strong tool to build model of time series that is accurate enough to be applied successfully in prediction task. In [11] a updated GMDH algorithm is used for medical image recognition. It is applied to medical image analysis of cancer of the liver. In this paper authors propose a novel feedback loop application in GMDH algorithm. In this approach, to organize the neural network architecture, a two types of neurons are applied: the polynomial type and the radial basis function (RBF)-type. It is shown that the feedback GMDH-type neural network algorithm is an accurate and a useful method for the nonlinear system identification. In this paper [12] GMDH (Group Method of Data Handling) has been applied for the identification of a mathematical model with many input variables. Proposed a Neuro-fuzzy GMDH model, adopting Gaussian radial basis functions (GRBF) as both a simplified fuzzy reasoning model and as a three-layered neural network. The Neuro-fuzzy GMDH algorithm is used for prediction of air pollution data. The authors compared the prediction accuracy of Neuro-fuzzy GMDH and Multi-Layer Perceptron (MLP). 3. Description of the algorithm Figure 1 shows the most general scheme of the algorithm, which illustrates how to achieve a sufficient complexity of the model. Algorithm receives a set of input variables (hereinafter referred to as arguments). In the first layer the selection of the arguments is made. The selected arguments should allow to build a model that will meet internal or external criterion. Internal criterion allows the constructed model to pass to the next layer, whereas the external criterion defines the requirements for the final model that is suitable for forecasting. Both of these criteria are selected in the learning process. Figure 1. Overall scheme of combinatorial algorithm, source [1], p. 41 Wiliński [1] has modified combinatorial algorithm and divided it into two algorithms. These algorithms have different objectives and with varying degrees they use historical data. The first algorithm, in the first layer, for each pair of arguments, calculates a linear regression model: ẑ i,j =a 0 +a 1 x i +a 2 x j. (2)

4 20 Michał Zabłocki Variables x i and x j are the time series of explanatory variables X, where i and j are the numbers of these series, but i j. Those time series have the same length but different time shifts regarding to the current moment. Neurons of the first layer examine the pairs of arguments by selection of the numerical coefficients vector A ij = [a 0, a 1, a 2 ], such that A ij = X ij \ y, (3) they are searching for the biggest convergence of constructed models with the vector of variable y. The left-division operator "\" solves the equation describing the explanatory variable y = A ij X ij. (4) This algorithm uses information gathered in the data matrix in such a way that the last element of the vector y corresponds to the value measured at the current time (or time taken as this one). The elements of X ij are shifted at least hp periods backward with respect to y. This prevents the use of information gathered in arguments from at least hp last moments. It allows to build the forecast for these moments. In the next step the vector of coefficients A is used to construct the prediction on hp periods ahead by substituting X ij shifted backward for X ij shifted by hp periods forward into equation (4). Hence, vector y falls in the area of the simulated results, that the real values at the present moment are still unknown. Thus, using historical data, it is possible to check which of the regression models gives the best estimate of the period considered. By comparing the model ẑ or just his predictive section ẑhp with the vector of explanatory variable y we can calculate the accuracy of prediction and select models, which will go to the next layer. In subsequent layers, the structure of regression models is described using the following formula: wˆ i, j,ii,jj,iii,jjj,... = a0 +a1xi +a2x j +a3zˆ ii +a4zˆ jj +a5vˆ iii +a6vˆ jjj +..., (5) where: i, j = 1, 2,..., M (number of explanatory variables), ii, jj, iii, jjjj = 1, 2,..., F y (number of models from previous layer), where i j, ii jj and iii jjj. The models arising in successive layers are extended by models that have been created in the previous layer. The second algorithm is a typical predictive algorithm. Using the structure chosen by the first algorithm and the most recent data, we build a predictive model and use it to make forecasts. In this algorithm, we can distinguish two phases. Network training phase in which the best model structure is chosen and phase where predictions are made: I. Network training phase (It uses the first algorithm and historical data, i.e., shifted backward from the current moment of at least 2 * hp periods.) 1. Calculation of numerical coefficients of polynomial regression models for the time shifted backward by hp periods from current moment. 2. Use obtained in step 1 numerical coefficients of models to build the forecasts of the explanatory variable. 3. Calculation of discrepancies between predicted and real values of observed variable according to the established criteria. 4. Selection of the model structure that will be used for forecasting. II. Prediction phase (In this phase, used data ends in the present moment.) 5. By use of model, built with the structure selected in the step 4, is performed prediction that can be implemented in the investment strategy. 6. After passing predictive horizon time period, on which forecast was built, the accuracy of reality reproduction is calculated for the model. 7. The structure of the model selected in step 4 is verified.

5 Agent-based distributed time series forecasting system The three modules of system based on GMDH algorithm The forecasting system presented in this article can work in one of three operating modes in order to build the forecast. In the first mode calculations are done with the involvement of three agents - server and two clients. In this mode, an investment decision is made by calculating two predictions for two different predictive horizon and comparing them with each other. Agent server compares the results received from the employed agents and computes the decision. The decision is positive (do action) if the forecast return is consistent, otherwise the decision is negative (abstain from action). In the second mode calculations are done with the involvement of four agents - server and three clients. In this mode three forecasts for the same predictive horizon are made. The forecasts differ in selection criteria applied to forecasting models selection. The following selection criteria were applied: a) the criterion of the best prediction made on the model, b) the criterion of the best representation of reality, and, c) the minimax criterion. The decision is made by a server agent as a result of comparing the three received forecasts. If all three returned forecast are consistent, the decision is positive, otherwise taking action on the basis of prediction is not recommended. The third mode assumes the maximum dispersion of calculations in order to obtain a decision as quick as possible. In this mode, there is no direct constraints of the number of agents calculating forecast, although the actual number of workers used to build the forecast depends on the number of explanatory variables. Each agent receives the same portion of the data on which performs calculations. The output decision is taken on the basis of the forecast, calculated for the same prediction horizon. 5. Research and calculations The extensive experiments were performed to examine the performance of presented in this article approach to the time series forecasting. Collected testing data consist of four time series representing OHLC 1 (15 minute time period 2 ) candles of EUR/USD currency pair. Each time series contains time samples on the period of time from 1 June 2009 at 03:15 to 13 March 2012 at 15:00. They were obtained through BOSSAFX trading platform which can be downloaded from Presented in the next three sections results of conducted experiments reveal the performance of presented system. In those experiments the system continuously has produced investment decisions. The starting point was selected (by "hit and miss") from the data set and were considered as a current moment. All values of explanatory variable that occur before that point were considered as a historical data. All values occurring after the selected point were considered as an unknown future values of the explanatory variables. After performing forecast and evaluating the results, the starting point was moved forward by hp candles, and the process repeated. In this way, the effectiveness of forecasts could be examined over a longer period. In application to real world, the system would had to wait for a period of time equal to the length of the prediction horizon in order to evaluate the accuracy of the forecast. As an indicators of the prediction quality the following measures were used: the cumulative sum of profits: 1 Open, High, Low, Close the values of variable in chosen time period. 2 In this period of time, by every 15 minutes, were taken measurements of OHLC value of EUR/USD currency pair.

6 22 Michał Zabłocki where: dz model return and dy - explanatory variable return. the maximum value of the prediction error: the mean value of prediction errors: the number of false prediction returns. 6. Testing forecasting performance, depending on the operation mode In this study the following parameters were adopted: vector length 50 candles, time study - 30 candles, the first candle had number 1004 and correspond to the date and time :00, 7 models always were passed to the next layer, prediction horizon was equal hp = 1, the forecasting model was always made by use of the best structure of the sixth layer. The study clearly shows (figures 2 and 3) that mode 2 gives the best forecasting results. The value of the cumulated sum of profits at the end of the study period reach a value of 0,019, which was a surprising result, because none of other modes were able to achieve similar result during the test (when this quality indicator was the highest). Mode 1 had the same number of correct decisions as the mode 2 and had similar values of average prediction error. However, other modes had smaller the maximum value of the prediction error. This reveal the fact that wrong decision can result in large losses. Mode 0 was the weakest strategy. The worst result of this strategy could be caused by the fact that the forecast for longer horizon could have bad influence on final decisions. The forecast for long horizon is usually burdened with greater prediction error. The system is misled even when the forecast for shorter horizon is correct. However, it is surprising that this mode had the lowest value of the average prediction error among all modes. In this case, can be postulated that it is the safest strategy that continues to generate profits. Figure 2. The cumulative sum of profits for the subsequent candles received by prediction on one candle ahead and using three modes of operation

7 Agent-based distributed time series forecasting system 23 Figure 3. The values of prediction errors made in successive candles received by prediction on one candle ahead and using three modes of operation. 7. Testing forecasting performance, depending on the method of models selection In this study the following parameters were adopted: vector length 50 candles, time study - 30 candles, the first candle had number 1400 and correspond to the date and time :00, 7 models always were passed to the next layer, prediction horizon was equal hp = 1, the forecasting model was always made by use of the best structure of the sixth layer. In the experiment the forecasting system with the use of mode 2 were applied, three computing agents were used, the upper limit of models number is calculated as follow: Fv= ( m(m1) 2 )( Fy(Fy1) 2 ) L1, (3) where m is the number of explanatory variables, and L is the number of layer. The basic quality criteria for the models was the average of all the model quality values obtained in the first layer. This value was further propagated to the criteria of next layers according to the formula: CR i =CR (1i 0.1) i1, (4) where i = 2,..., 6 is a consecutive numbers of layers. The quality of the prediction model was determined based on the accuracy of mapping the actual course of observed variable. The following methods of models selection was tested: the choice of F y first models that meet the criterion (s.m. 1), roulette method (s.m. 2), the selection of F y the best models (s.m. 3), ranking method (s.m. 4), tournament method (s.m. 5), random selection (s.m. 6). The method based on the selection of F y the best models (s.m. 3) turned out to be the best selection method (see figure 4 and 5). This method extends in some way the method consisting in selection of first F y models that meet the criteria (s.m. 1). The winning method ensures that the next layer will also include (or rather only) the best models. This is the socalled elitism, which Ivahnenko (creator of the concept of GMDH method) did not want to use. Nevertheless, it appears that the best models obtain the best results. However, it is good to focus on the random models selection (s.m. 6) results. This method confirms the idea that worse models can in successive layers provide better models.

8 24 Michał Zabłocki Of course, in this case the choice of model that goes to the next layer, is a pure coincidence, and it is hard to determine the quality of models, that passed to the next layer. However, we can assume that not only the best models passed to next layer. The results of simulations, in which were used the selection methods borrowed from the theory of genetic algorithms (s.m. 2, s.m. 4, s.m. 5), show that those methods have a tendency to causing a single, relatively large, errors. In the consequence the cumulative sum of the profit is significantly reduced. However, they do not outperform other methods in terms of the average value of the prediction error. Moreover, they showed a better forecasts accuracy than the other methods. The tournament method (s.m. 5) were the leading among those methods. Figure 4. The cumulative sum of profits for the subsequent candles received by prediction on one candle ahead and six selection methods Figure 5. The values of prediction errors made in successive candles received by prediction on one candle ahead and six selection methods

9 Agent-based distributed time series forecasting system Testing forecasting performance, depending on the adopted prediction horizon and the method of models selection In this study the following parameters were adopted: vector length 50 candles, time study - 30 candles, the first candle had number 1400 and correspond to the date and time :00, 7 models always were passed to the next layer, prediction horizon was equal hp = 1, the forecasting model was always made by use of the best structure of the sixth layer. The study was conducted with the use of combinied modes 1 and 2: four computing agents were used, the upper limit of models number can be calculated by use of equation (3). the basic criterion for the quality of the model were the same as in previous experiment. In this study were tested prediction horizon on two and three candles forward, and the following methods of models selection was tested: the choice of F y first models that meet the criterion, roulette method, the selection of F y the best models, ranking method, tournament method, random selection. The method based on the selection of F y the best models (s.m. 3) outperforms other methods (see figure 6 and 7). This method achieved the highest results for tested prediction horizons. It has a small value of the maximum prediction error and the average prediction error. Roulette (s.m. 2) was only slightly less effective. It reaches 70% of maximum profit for studied prediction horizons. It had more than 63% in the effectiveness of the decision making, which is a better result than the s.m. 3 method. Unfortunately, s.m. 2 achieved also the highest values of maximum prediction error and the mean prediction error. Other selection methods based on the theory of genetic algorithms had much worst results. The worst results had the tournament method (s.m. 5) which had poor results (53-64%). Figure 6. The cumulated sum of profits for the subsequent candles expressed in percentage of maximum profit received by prediction on two candle ahead and using six selection methods (s. m.)

10 26 Michał Zabłocki Figure 7. The cumulated sum of profits for the subsequent candles expressed in percentage of maximum profit received by prediction on three candle ahead and using six selection methods. 9. Conclusions The experimental results reveals that the system is able to produce a reliable investment decisions. As shown in conducted analysis, in spite of the fact that the system makes mistakes in individual forecasts the cumulative sum of profit is positive. Presented in this article, time series forecasting system has great potential. These studies demonstrate that the system can be deployed on the trading platform to automate process of the prediction of financial markets. Described in this paper distributed approach extends used by Wiliński [1] GMDH method. It achieves even better results and gives the ability to benefit from distributed computer resources. The proposed changes had a positive influence on the obtained results. Further research aims at improving an individual forecast and to develop more sophisticated investment strategies. The use of more sophisticated statistical methods can result in development of better prognostic indicators, which can improve the accuracy of the prediction. References [1] Wiliński A., GMDH metody grupowania argumentów w zadaniach zautomatyzowanej predykcji zachowań rynków finansowych, Warszawa - Szczecin 2009, 278 s., ISBN [2] GMDH - Group Method of Data Handling, [3] Ivakhnenko A., Ivakhnenko G., Problems of Further Development of the Group Method of Data Handling Algorithms, Part I. Pattern Recognition and Image Analysis vol.10 No.2, pp , [4] Ivakhnenko A., Ivakhnenko G.,Mueller J., Self-organization of Neural Network with Active Neurons, Part I. Pattern Recognition and Image Analysis, vol.4 No.2, pp , [5] Ivakhnenko A.G., Ivakhnenko G.A., Andrienko N.M. Inductive Computer Advisor for Current forecasting of Ukraine's Macroeconomy, Systems Analysis Modelling Simulation, 22, no.1, 1998.

11 Agent-based distributed time series forecasting system 27 [6] Ivakhnenko G.A., Model-Free Analogues As Active Neurons for Neural Networks Self- Organization, Control Systems and Computers, no.2, p , [7] Madala H.R., Ivakhnenko A.G., Inductive Learning Algorithms for Complex Systems Modelling. CRC Press Inc.. Boca Raton, Ann Arbor, London, Tokyo, ISBN: , 1994 [8] Rogoza, V., Zabłocki, M., Grid computing and Cloud computing in scope of JADE and OWL based Semantic Agents A Survey, Przegląd Elektrotechniczny, 90, 2/2014, ISSN [9] Bouška J., Kordík P., Time Series Prediction by means of GMDH Analogues Complexing and GAME (Paper in Conference Proceedings), In IWIM International Workshop on Inductive Modelling. Praha: Czech Technical University in Prague, 2007, p ISBN [10] Samsudin R., Saad P., Shabri A., A hybrid least squares support vector machines and GMDH approach for river flow forecasting, Hydrol. Earth Syst. Sci. Discuss., 7, , doi: /hessd , 2010 [11] Kondo T., Kondo Ch., Takao S.,Ueno J.,Feedback GMDH-type neural network algorithm and its application to medical image analysis of cancer of the liver, Artificial Life and Robotics, Volume 15, Issue 3, p , 2010 [12] Yousefpour A., Ahmadpour Z., The prediction of air pollution by using Neuro-fuzzy GMDH, The Journal of Mathematics and Computer Science, Vol.2, No.3,p , 2011.

A FUZZY NEURAL NETWORK MODEL FOR FORECASTING STOCK PRICE

A FUZZY NEURAL NETWORK MODEL FOR FORECASTING STOCK PRICE A FUZZY NEURAL NETWORK MODEL FOR FORECASTING STOCK PRICE Li Sheng Institute of intelligent information engineering Zheiang University Hangzhou, 3007, P. R. China ABSTRACT In this paper, a neural network-driven

More information

FEEDBACK GMDH-TYPE NEURAL NETWORK AND ITS APPLICATION TO MEDICAL IMAGE ANALYSIS OF LIVER CANCER. Tadashi Kondo and Junji Ueno

FEEDBACK GMDH-TYPE NEURAL NETWORK AND ITS APPLICATION TO MEDICAL IMAGE ANALYSIS OF LIVER CANCER. Tadashi Kondo and Junji Ueno International Journal of Innovative Computing, Information and Control ICIC International c 2012 ISSN 1349-4198 Volume 8, Number 3(B), March 2012 pp. 2285 2300 FEEDBACK GMDH-TYPE NEURAL NETWORK AND ITS

More information

Capabilities and Prospects of Inductive Modeling Volodymyr STEPASHKO

Capabilities and Prospects of Inductive Modeling Volodymyr STEPASHKO Capabilities and Prospects of Inductive Modeling Volodmr STEPASHKO Prof., Dr. Sci., Head of Department INFORMATION TECHNOLOGIES FOR INDUCTIVE MODELING International Research and Training Centre of the

More information

Robust Pareto Design of GMDH-type Neural Networks for Systems with Probabilistic Uncertainties

Robust Pareto Design of GMDH-type Neural Networks for Systems with Probabilistic Uncertainties . Hybrid GMDH-type algorithms and neural networks Robust Pareto Design of GMDH-type eural etworks for Systems with Probabilistic Uncertainties. ariman-zadeh, F. Kalantary, A. Jamali, F. Ebrahimi Department

More information

GMDH-type Neural Networks with a Feedback Loop and their Application to the Identification of Large-spatial Air Pollution Patterns.

GMDH-type Neural Networks with a Feedback Loop and their Application to the Identification of Large-spatial Air Pollution Patterns. GMDH-type Neural Networks with a Feedback Loop and their Application to the Identification of Large-spatial Air Pollution Patterns. Tadashi Kondo 1 and Abhijit S.Pandya 2 1 School of Medical Sci.,The Univ.of

More information

22/04/2014. Economic Research

22/04/2014. Economic Research 22/04/2014 Economic Research Forecasting Models for Exchange Rate Tuesday, April 22, 2014 The science of prognostics has been going through a rapid and fruitful development in the past decades, with various

More information

ADAPTIVE NEURO-FUZZY INFERENCE SYSTEMS

ADAPTIVE NEURO-FUZZY INFERENCE SYSTEMS ADAPTIVE NEURO-FUZZY INFERENCE SYSTEMS RBFN and TS systems Equivalent if the following hold: Both RBFN and TS use same aggregation method for output (weighted sum or weighted average) Number of basis functions

More information

Forecasting River Flow in the USA: A Comparison between Auto-Regression and Neural Network Non-Parametric Models

Forecasting River Flow in the USA: A Comparison between Auto-Regression and Neural Network Non-Parametric Models Journal of Computer Science 2 (10): 775-780, 2006 ISSN 1549-3644 2006 Science Publications Forecasting River Flow in the USA: A Comparison between Auto-Regression and Neural Network Non-Parametric Models

More information

SARIMA-ELM Hybrid Model for Forecasting Tourist in Nepal

SARIMA-ELM Hybrid Model for Forecasting Tourist in Nepal Volume-03 Issue-07 July-2018 ISSN: 2455-3085 (Online) www.rrjournals.com [UGC Listed Journal] SARIMA-ELM Hybrid Model for Forecasting Tourist in Nepal *1 Kadek Jemmy Waciko & 2 Ismail B *1 Research Scholar,

More information

CHAPTER 6 CONCLUSION AND FUTURE SCOPE

CHAPTER 6 CONCLUSION AND FUTURE SCOPE CHAPTER 6 CONCLUSION AND FUTURE SCOPE 146 CHAPTER 6 CONCLUSION AND FUTURE SCOPE 6.1 SUMMARY The first chapter of the thesis highlighted the need of accurate wind forecasting models in order to transform

More information

ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD

ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD WHAT IS A NEURAL NETWORK? The simplest definition of a neural network, more properly referred to as an 'artificial' neural network (ANN), is provided

More information

How Accurate is My Forecast?

How Accurate is My Forecast? How Accurate is My Forecast? Tao Hong, PhD Utilities Business Unit, SAS 15 May 2012 PLEASE STAND BY Today s event will begin at 11:00am EDT The audio portion of the presentation will be heard through your

More information

Analysis of Fast Input Selection: Application in Time Series Prediction

Analysis of Fast Input Selection: Application in Time Series Prediction Analysis of Fast Input Selection: Application in Time Series Prediction Jarkko Tikka, Amaury Lendasse, and Jaakko Hollmén Helsinki University of Technology, Laboratory of Computer and Information Science,

More information

Neural Networks and the Back-propagation Algorithm

Neural Networks and the Back-propagation Algorithm Neural Networks and the Back-propagation Algorithm Francisco S. Melo In these notes, we provide a brief overview of the main concepts concerning neural networks and the back-propagation algorithm. We closely

More information

Short Term Load Forecasting Using Multi Layer Perceptron

Short Term Load Forecasting Using Multi Layer Perceptron International OPEN ACCESS Journal Of Modern Engineering Research (IJMER) Short Term Load Forecasting Using Multi Layer Perceptron S.Hema Chandra 1, B.Tejaswini 2, B.suneetha 3, N.chandi Priya 4, P.Prathima

More information

DEVS Simulation of Spiking Neural Networks

DEVS Simulation of Spiking Neural Networks DEVS Simulation of Spiking Neural Networks Rene Mayrhofer, Michael Affenzeller, Herbert Prähofer, Gerhard Höfer, Alexander Fried Institute of Systems Science Systems Theory and Information Technology Johannes

More information

Integrated Electricity Demand and Price Forecasting

Integrated Electricity Demand and Price Forecasting Integrated Electricity Demand and Price Forecasting Create and Evaluate Forecasting Models The many interrelated factors which influence demand for electricity cannot be directly modeled by closed-form

More information

Artificial Neural Networks The Introduction

Artificial Neural Networks The Introduction Artificial Neural Networks The Introduction 01001110 01100101 01110101 01110010 01101111 01101110 01101111 01110110 01100001 00100000 01110011 01101011 01110101 01110000 01101001 01101110 01100001 00100000

More information

LONG - TERM INDUSTRIAL LOAD FORECASTING AND PLANNING USING NEURAL NETWORKS TECHNIQUE AND FUZZY INFERENCE METHOD ABSTRACT

LONG - TERM INDUSTRIAL LOAD FORECASTING AND PLANNING USING NEURAL NETWORKS TECHNIQUE AND FUZZY INFERENCE METHOD ABSTRACT LONG - TERM NDUSTRAL LOAD FORECASTNG AND PLANNNG USNG NEURAL NETWORKS TECHNQUE AND FUZZY NFERENCE METHOD M. A. Farahat Zagazig University, Zagazig, Egypt ABSTRACT Load forecasting plays a dominant part

More information

COMPARISON OF CLEAR-SKY MODELS FOR EVALUATING SOLAR FORECASTING SKILL

COMPARISON OF CLEAR-SKY MODELS FOR EVALUATING SOLAR FORECASTING SKILL COMPARISON OF CLEAR-SKY MODELS FOR EVALUATING SOLAR FORECASTING SKILL Ricardo Marquez Mechanical Engineering and Applied Mechanics School of Engineering University of California Merced Carlos F. M. Coimbra

More information

Natural Gas Prediction Using The Group Method of Data Handling

Natural Gas Prediction Using The Group Method of Data Handling Natural Gas Prediction Using The Group Method of Data Handling James C. Howland III Computer Information Systems College of Technical Sciences Montana State University NORTHERN Havre, Montana, USA howland@msun.edu

More information

RAINFALL RUNOFF MODELING USING SUPPORT VECTOR REGRESSION AND ARTIFICIAL NEURAL NETWORKS

RAINFALL RUNOFF MODELING USING SUPPORT VECTOR REGRESSION AND ARTIFICIAL NEURAL NETWORKS CEST2011 Rhodes, Greece Ref no: XXX RAINFALL RUNOFF MODELING USING SUPPORT VECTOR REGRESSION AND ARTIFICIAL NEURAL NETWORKS D. BOTSIS1 1, P. LATINOPOULOS 2 and K. DIAMANTARAS 3 1&2 Department of Civil

More information

Improved Holt Method for Irregular Time Series

Improved Holt Method for Irregular Time Series WDS'08 Proceedings of Contributed Papers, Part I, 62 67, 2008. ISBN 978-80-7378-065-4 MATFYZPRESS Improved Holt Method for Irregular Time Series T. Hanzák Charles University, Faculty of Mathematics and

More information

About Nnergix +2, More than 2,5 GW forecasted. Forecasting in 5 countries. 4 predictive technologies. More than power facilities

About Nnergix +2, More than 2,5 GW forecasted. Forecasting in 5 countries. 4 predictive technologies. More than power facilities About Nnergix +2,5 5 4 +20.000 More than 2,5 GW forecasted Forecasting in 5 countries 4 predictive technologies More than 20.000 power facilities Nnergix s Timeline 2012 First Solar Photovoltaic energy

More information

Choosing Variables with a Genetic Algorithm for Econometric models based on Neural Networks learning and adaptation.

Choosing Variables with a Genetic Algorithm for Econometric models based on Neural Networks learning and adaptation. Choosing Variables with a Genetic Algorithm for Econometric models based on Neural Networks learning and adaptation. Daniel Ramírez A., Israel Truijillo E. LINDA LAB, Computer Department, UNAM Facultad

More information

Application of Artificial Neural Networks in Evaluation and Identification of Electrical Loss in Transformers According to the Energy Consumption

Application of Artificial Neural Networks in Evaluation and Identification of Electrical Loss in Transformers According to the Energy Consumption Application of Artificial Neural Networks in Evaluation and Identification of Electrical Loss in Transformers According to the Energy Consumption ANDRÉ NUNES DE SOUZA, JOSÉ ALFREDO C. ULSON, IVAN NUNES

More information

Evolutionary Functional Link Interval Type-2 Fuzzy Neural System for Exchange Rate Prediction

Evolutionary Functional Link Interval Type-2 Fuzzy Neural System for Exchange Rate Prediction Evolutionary Functional Link Interval Type-2 Fuzzy Neural System for Exchange Rate Prediction 3. Introduction Currency exchange rate is an important element in international finance. It is one of the chaotic,

More information

ANN based techniques for prediction of wind speed of 67 sites of India

ANN based techniques for prediction of wind speed of 67 sites of India ANN based techniques for prediction of wind speed of 67 sites of India Paper presentation in Conference on Large Scale Grid Integration of Renewable Energy in India Authors: Parul Arora Prof. B.K Panigrahi

More information

Data Mining Part 5. Prediction

Data Mining Part 5. Prediction Data Mining Part 5. Prediction 5.5. Spring 2010 Instructor: Dr. Masoud Yaghini Outline How the Brain Works Artificial Neural Networks Simple Computing Elements Feed-Forward Networks Perceptrons (Single-layer,

More information

Artificial Neural Networks (ANN) Xiaogang Su, Ph.D. Department of Mathematical Science University of Texas at El Paso

Artificial Neural Networks (ANN) Xiaogang Su, Ph.D. Department of Mathematical Science University of Texas at El Paso Artificial Neural Networks (ANN) Xiaogang Su, Ph.D. Department of Mathematical Science University of Texas at El Paso xsu@utep.edu Fall, 2018 Outline Introduction A Brief History ANN Architecture Terminology

More information

Predict Time Series with Multiple Artificial Neural Networks

Predict Time Series with Multiple Artificial Neural Networks , pp. 313-324 http://dx.doi.org/10.14257/ijhit.2016.9.7.28 Predict Time Series with Multiple Artificial Neural Networks Fei Li 1, Jin Liu 1 and Lei Kong 2,* 1 College of Information Engineering, Shanghai

More information

One-Hour-Ahead Load Forecasting Using Neural Network

One-Hour-Ahead Load Forecasting Using Neural Network IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 17, NO. 1, FEBRUARY 2002 113 One-Hour-Ahead Load Forecasting Using Neural Network Tomonobu Senjyu, Member, IEEE, Hitoshi Takara, Katsumi Uezato, and Toshihisa Funabashi,

More information

Probabilistic Energy Forecasting

Probabilistic Energy Forecasting Probabilistic Energy Forecasting Moritz Schmid Seminar Energieinformatik WS 2015/16 ^ KIT The Research University in the Helmholtz Association www.kit.edu Agenda Forecasting challenges Renewable energy

More information

Short Term Load Forecasting Based Artificial Neural Network

Short Term Load Forecasting Based Artificial Neural Network Short Term Load Forecasting Based Artificial Neural Network Dr. Adel M. Dakhil Department of Electrical Engineering Misan University Iraq- Misan Dr.adelmanaa@gmail.com Abstract Present study develops short

More information

Advanced Weather Technology

Advanced Weather Technology Advanced Weather Technology Tuesday, October 16, 2018, 1:00 PM 2:00 PM PRESENTED BY: Gary Pokodner, FAA WTIC Program Manager Agenda Overview Augmented reality mobile application Crowd Sourcing Visibility

More information

Financial Risk and Returns Prediction with Modular Networked Learning

Financial Risk and Returns Prediction with Modular Networked Learning arxiv:1806.05876v1 [cs.lg] 15 Jun 2018 Financial Risk and Returns Prediction with Modular Networked Learning Carlos Pedro Gonçalves June 18, 2018 University of Lisbon, Instituto Superior de Ciências Sociais

More information

Engineering Part IIB: Module 4F10 Statistical Pattern Processing Lecture 6: Multi-Layer Perceptrons I

Engineering Part IIB: Module 4F10 Statistical Pattern Processing Lecture 6: Multi-Layer Perceptrons I Engineering Part IIB: Module 4F10 Statistical Pattern Processing Lecture 6: Multi-Layer Perceptrons I Phil Woodland: pcw@eng.cam.ac.uk Michaelmas 2012 Engineering Part IIB: Module 4F10 Introduction In

More information

CHAPTER 7 CONCLUSION AND FUTURE WORK

CHAPTER 7 CONCLUSION AND FUTURE WORK 159 CHAPTER 7 CONCLUSION AND FUTURE WORK 7.1 INTRODUCTION Nonlinear time series analysis is an important area of research in all fields of science and engineering. It is an important component of operations

More information

Pattern Matching and Neural Networks based Hybrid Forecasting System

Pattern Matching and Neural Networks based Hybrid Forecasting System Pattern Matching and Neural Networks based Hybrid Forecasting System Sameer Singh and Jonathan Fieldsend PA Research, Department of Computer Science, University of Exeter, Exeter, UK Abstract In this paper

More information

A Hybrid Model of Wavelet and Neural Network for Short Term Load Forecasting

A Hybrid Model of Wavelet and Neural Network for Short Term Load Forecasting International Journal of Electronic and Electrical Engineering. ISSN 0974-2174, Volume 7, Number 4 (2014), pp. 387-394 International Research Publication House http://www.irphouse.com A Hybrid Model of

More information

Forecasting demand in the National Electricity Market. October 2017

Forecasting demand in the National Electricity Market. October 2017 Forecasting demand in the National Electricity Market October 2017 Agenda Trends in the National Electricity Market A review of AEMO s forecasting methods Long short-term memory (LSTM) neural networks

More information

WEATHER DEPENENT ELECTRICITY MARKET FORECASTING WITH NEURAL NETWORKS, WAVELET AND DATA MINING TECHNIQUES. Z.Y. Dong X. Li Z. Xu K. L.

WEATHER DEPENENT ELECTRICITY MARKET FORECASTING WITH NEURAL NETWORKS, WAVELET AND DATA MINING TECHNIQUES. Z.Y. Dong X. Li Z. Xu K. L. WEATHER DEPENENT ELECTRICITY MARKET FORECASTING WITH NEURAL NETWORKS, WAVELET AND DATA MINING TECHNIQUES Abstract Z.Y. Dong X. Li Z. Xu K. L. Teo School of Information Technology and Electrical Engineering

More information

To Predict Rain Fall in Desert Area of Rajasthan Using Data Mining Techniques

To Predict Rain Fall in Desert Area of Rajasthan Using Data Mining Techniques To Predict Rain Fall in Desert Area of Rajasthan Using Data Mining Techniques Peeyush Vyas Asst. Professor, CE/IT Department of Vadodara Institute of Engineering, Vadodara Abstract: Weather forecasting

More information

1. Introduction. 2. Artificial Neural Networks and Fuzzy Time Series

1. Introduction. 2. Artificial Neural Networks and Fuzzy Time Series 382 IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.9, September 2008 A Comparative Study of Neural-Network & Fuzzy Time Series Forecasting Techniques Case Study: Wheat

More information

Outline Introduction OLS Design of experiments Regression. Metamodeling. ME598/494 Lecture. Max Yi Ren

Outline Introduction OLS Design of experiments Regression. Metamodeling. ME598/494 Lecture. Max Yi Ren 1 / 34 Metamodeling ME598/494 Lecture Max Yi Ren Department of Mechanical Engineering, Arizona State University March 1, 2015 2 / 34 1. preliminaries 1.1 motivation 1.2 ordinary least square 1.3 information

More information

SOFTWARE FOR THE ANALYSIS OF THE COMPLETE FACTORIAL EXPERIMENT IN MECHANICAL ENGINEERING

SOFTWARE FOR THE ANALYSIS OF THE COMPLETE FACTORIAL EXPERIMENT IN MECHANICAL ENGINEERING Nonconventional Technologies Review Romania, December, 2013 2013 Romanian Association of Nonconventional Technologies SOFTWARE FOR THE ANALYSIS OF THE COMPLETE FACTORIAL EXPERIMENT IN MECHANICAL ENGINEERING

More information

FORECASTING OF ECONOMIC QUANTITIES USING FUZZY AUTOREGRESSIVE MODEL AND FUZZY NEURAL NETWORK

FORECASTING OF ECONOMIC QUANTITIES USING FUZZY AUTOREGRESSIVE MODEL AND FUZZY NEURAL NETWORK FORECASTING OF ECONOMIC QUANTITIES USING FUZZY AUTOREGRESSIVE MODEL AND FUZZY NEURAL NETWORK Dusan Marcek Silesian University, Institute of Computer Science Opava Research Institute of the IT4Innovations

More information

Machine Learning to Automatically Detect Human Development from Satellite Imagery

Machine Learning to Automatically Detect Human Development from Satellite Imagery Technical Disclosure Commons Defensive Publications Series April 24, 2017 Machine Learning to Automatically Detect Human Development from Satellite Imagery Matthew Manolides Follow this and additional

More information

MODELLING ENERGY DEMAND FORECASTING USING NEURAL NETWORKS WITH UNIVARIATE TIME SERIES

MODELLING ENERGY DEMAND FORECASTING USING NEURAL NETWORKS WITH UNIVARIATE TIME SERIES MODELLING ENERGY DEMAND FORECASTING USING NEURAL NETWORKS WITH UNIVARIATE TIME SERIES S. Cankurt 1, M. Yasin 2 1&2 Ishik University Erbil, Iraq 1 s.cankurt@ishik.edu.iq, 2 m.yasin@ishik.edu.iq doi:10.23918/iec2018.26

More information

Artificial Neural Networks Examination, March 2002

Artificial Neural Networks Examination, March 2002 Artificial Neural Networks Examination, March 2002 Instructions There are SIXTY questions (worth up to 60 marks). The exam mark (maximum 60) will be added to the mark obtained in the laborations (maximum

More information

NONLINEAR CLASSIFICATION AND REGRESSION. J. Elder CSE 4404/5327 Introduction to Machine Learning and Pattern Recognition

NONLINEAR CLASSIFICATION AND REGRESSION. J. Elder CSE 4404/5327 Introduction to Machine Learning and Pattern Recognition NONLINEAR CLASSIFICATION AND REGRESSION Nonlinear Classification and Regression: Outline 2 Multi-Layer Perceptrons The Back-Propagation Learning Algorithm Generalized Linear Models Radial Basis Function

More information

POWER SYSTEM DYNAMIC SECURITY ASSESSMENT CLASSICAL TO MODERN APPROACH

POWER SYSTEM DYNAMIC SECURITY ASSESSMENT CLASSICAL TO MODERN APPROACH Abstract POWER SYSTEM DYNAMIC SECURITY ASSESSMENT CLASSICAL TO MODERN APPROACH A.H.M.A.Rahim S.K.Chakravarthy Department of Electrical Engineering K.F. University of Petroleum and Minerals Dhahran. Dynamic

More information

Introduction to Natural Computation. Lecture 9. Multilayer Perceptrons and Backpropagation. Peter Lewis

Introduction to Natural Computation. Lecture 9. Multilayer Perceptrons and Backpropagation. Peter Lewis Introduction to Natural Computation Lecture 9 Multilayer Perceptrons and Backpropagation Peter Lewis 1 / 25 Overview of the Lecture Why multilayer perceptrons? Some applications of multilayer perceptrons.

More information

Forecasting of Rain Fall in Mirzapur District, Uttar Pradesh, India Using Feed-Forward Artificial Neural Network

Forecasting of Rain Fall in Mirzapur District, Uttar Pradesh, India Using Feed-Forward Artificial Neural Network International Journal of Engineering Science Invention ISSN (Online): 2319 6734, ISSN (Print): 2319 6726 Volume 2 Issue 8ǁ August. 2013 ǁ PP.87-93 Forecasting of Rain Fall in Mirzapur District, Uttar Pradesh,

More information

CSC321 Lecture 2: Linear Regression

CSC321 Lecture 2: Linear Regression CSC32 Lecture 2: Linear Regression Roger Grosse Roger Grosse CSC32 Lecture 2: Linear Regression / 26 Overview First learning algorithm of the course: linear regression Task: predict scalar-valued targets,

More information

MODEL FOR DISTRIBUTION OF WAGES

MODEL FOR DISTRIBUTION OF WAGES 19th Applications of Mathematics and Statistics in Economics AMSE 216 MODEL FOR DISTRIBUTION OF WAGES MICHAL VRABEC, LUBOŠ MAREK University of Economics, Prague, Faculty of Informatics and Statistics,

More information

Optimum Neural Network Architecture for Precipitation Prediction of Myanmar

Optimum Neural Network Architecture for Precipitation Prediction of Myanmar Optimum Neural Network Architecture for Precipitation Prediction of Myanmar Khaing Win Mar, Thinn Thu Naing Abstract Nowadays, precipitation prediction is required for proper planning and management of

More information

Neural Inversion Technology for reservoir property prediction from seismic data

Neural Inversion Technology for reservoir property prediction from seismic data Original article published in Russian in Nefteservice, March 2009 Neural Inversion Technology for reservoir property prediction from seismic data Malyarova Tatyana, Kopenkin Roman, Paradigm At the software

More information

International Journal of Scientific Research and Reviews

International Journal of Scientific Research and Reviews Research article Available online www.ijsrr.org ISSN: 2279 0543 International Journal of Scientific Research and Reviews Prediction of Compressive Strength of Concrete using Artificial Neural Network ABSTRACT

More information

Deep Feedforward Networks

Deep Feedforward Networks Deep Feedforward Networks Liu Yang March 30, 2017 Liu Yang Short title March 30, 2017 1 / 24 Overview 1 Background A general introduction Example 2 Gradient based learning Cost functions Output Units 3

More information

ARTIFICIAL NEURAL NETWORKS APPROACH IN MICROWAVE FILTER TUNING

ARTIFICIAL NEURAL NETWORKS APPROACH IN MICROWAVE FILTER TUNING Progress In Electromagnetics Research M, Vol. 13, 173 188, 2010 ARTIFICIAL NEURAL NETWORKS APPROACH IN MICROWAVE FILTER TUNING J. J. Michalski TeleMobile Electronics Ltd. Pomeranian Science and Technology

More information

ARTIFICIAL NEURAL NETWORKS گروه مطالعاتي 17 بهار 92

ARTIFICIAL NEURAL NETWORKS گروه مطالعاتي 17 بهار 92 ARTIFICIAL NEURAL NETWORKS گروه مطالعاتي 17 بهار 92 BIOLOGICAL INSPIRATIONS Some numbers The human brain contains about 10 billion nerve cells (neurons) Each neuron is connected to the others through 10000

More information

Machine Learning for Large-Scale Data Analysis and Decision Making A. Neural Networks Week #6

Machine Learning for Large-Scale Data Analysis and Decision Making A. Neural Networks Week #6 Machine Learning for Large-Scale Data Analysis and Decision Making 80-629-17A Neural Networks Week #6 Today Neural Networks A. Modeling B. Fitting C. Deep neural networks Today s material is (adapted)

More information

A SEASONAL FUZZY TIME SERIES FORECASTING METHOD BASED ON GUSTAFSON-KESSEL FUZZY CLUSTERING *

A SEASONAL FUZZY TIME SERIES FORECASTING METHOD BASED ON GUSTAFSON-KESSEL FUZZY CLUSTERING * No.2, Vol.1, Winter 2012 2012 Published by JSES. A SEASONAL FUZZY TIME SERIES FORECASTING METHOD BASED ON GUSTAFSON-KESSEL * Faruk ALPASLAN a, Ozge CAGCAG b Abstract Fuzzy time series forecasting methods

More information

Ocean Based Water Allocation Forecasts Using an Artificial Intelligence Approach

Ocean Based Water Allocation Forecasts Using an Artificial Intelligence Approach Ocean Based Water Allocation Forecasts Using an Artificial Intelligence Approach Khan S 1, Dassanayake D 2 and Rana T 2 1 Charles Sturt University and CSIRO Land and Water, School of Science and Tech,

More information

Data and prognosis for renewable energy

Data and prognosis for renewable energy The Hong Kong Polytechnic University Department of Electrical Engineering Project code: FYP_27 Data and prognosis for renewable energy by Choi Man Hin 14072258D Final Report Bachelor of Engineering (Honours)

More information

Neural Network Weight Space Symmetries Can Speed up Genetic Learning

Neural Network Weight Space Symmetries Can Speed up Genetic Learning Neural Network Weight Space Symmetries Can Speed up Genetic Learning ROMAN NERUDA Λ Institue of Computer Science Academy of Sciences of the Czech Republic P.O. Box 5, 187 Prague, Czech Republic tel: (4)665375,fax:(4)8585789

More information

EE-588 ADVANCED TOPICS IN NEURAL NETWORK

EE-588 ADVANCED TOPICS IN NEURAL NETWORK CUKUROVA UNIVERSITY DEPARTMENT OF ELECTRICAL&ELECTRONICS ENGINEERING EE-588 ADVANCED TOPICS IN NEURAL NETWORK THE PROJECT PROPOSAL AN APPLICATION OF NEURAL NETWORKS FOR WEATHER TEMPERATURE FORECASTING

More information

Explaining Results of Neural Networks by Contextual Importance and Utility

Explaining Results of Neural Networks by Contextual Importance and Utility Explaining Results of Neural Networks by Contextual Importance and Utility Kary FRÄMLING Dep. SIMADE, Ecole des Mines, 158 cours Fauriel, 42023 Saint-Etienne Cedex 2, FRANCE framling@emse.fr, tel.: +33-77.42.66.09

More information

UniResearch Ltd, University of Bergen, Bergen, Norway WinSim Ltd., Tonsberg, Norway {catherine,

UniResearch Ltd, University of Bergen, Bergen, Norway WinSim Ltd., Tonsberg, Norway {catherine, Improving an accuracy of ANN-based mesoscalemicroscale coupling model by data categorization: with application to wind forecast for offshore and complex terrain onshore wind farms. Alla Sapronova 1*, Catherine

More information

AN ARTIFICIAL NEURAL NETWORK MODEL FOR ROAD ACCIDENT PREDICTION: A CASE STUDY OF KHULNA METROPOLITAN CITY

AN ARTIFICIAL NEURAL NETWORK MODEL FOR ROAD ACCIDENT PREDICTION: A CASE STUDY OF KHULNA METROPOLITAN CITY Proceedings of the 4 th International Conference on Civil Engineering for Sustainable Development (ICCESD 2018), 9~11 February 2018, KUET, Khulna, Bangladesh (ISBN-978-984-34-3502-6) AN ARTIFICIAL NEURAL

More information

A Support Vector Regression Model for Forecasting Rainfall

A Support Vector Regression Model for Forecasting Rainfall A Support Vector Regression for Forecasting Nasimul Hasan 1, Nayan Chandra Nath 1, Risul Islam Rasel 2 Department of Computer Science and Engineering, International Islamic University Chittagong, Bangladesh

More information

Multitask Learning of Environmental Spatial Data

Multitask Learning of Environmental Spatial Data 9th International Congress on Environmental Modelling and Software Brigham Young University BYU ScholarsArchive 6th International Congress on Environmental Modelling and Software - Leipzig, Germany - July

More information

Artificial Intelligence (AI) Common AI Methods. Training. Signals to Perceptrons. Artificial Neural Networks (ANN) Artificial Intelligence

Artificial Intelligence (AI) Common AI Methods. Training. Signals to Perceptrons. Artificial Neural Networks (ANN) Artificial Intelligence Artificial Intelligence (AI) Artificial Intelligence AI is an attempt to reproduce intelligent reasoning using machines * * H. M. Cartwright, Applications of Artificial Intelligence in Chemistry, 1993,

More information

ARTIFICIAL NEURAL NETWORK WITH HYBRID TAGUCHI-GENETIC ALGORITHM FOR NONLINEAR MIMO MODEL OF MACHINING PROCESSES

ARTIFICIAL NEURAL NETWORK WITH HYBRID TAGUCHI-GENETIC ALGORITHM FOR NONLINEAR MIMO MODEL OF MACHINING PROCESSES International Journal of Innovative Computing, Information and Control ICIC International c 2013 ISSN 1349-4198 Volume 9, Number 4, April 2013 pp. 1455 1475 ARTIFICIAL NEURAL NETWORK WITH HYBRID TAGUCHI-GENETIC

More information

Balancing and Control of a Freely-Swinging Pendulum Using a Model-Free Reinforcement Learning Algorithm

Balancing and Control of a Freely-Swinging Pendulum Using a Model-Free Reinforcement Learning Algorithm Balancing and Control of a Freely-Swinging Pendulum Using a Model-Free Reinforcement Learning Algorithm Michail G. Lagoudakis Department of Computer Science Duke University Durham, NC 2778 mgl@cs.duke.edu

More information

AN INTRODUCTION TO NEURAL NETWORKS. Scott Kuindersma November 12, 2009

AN INTRODUCTION TO NEURAL NETWORKS. Scott Kuindersma November 12, 2009 AN INTRODUCTION TO NEURAL NETWORKS Scott Kuindersma November 12, 2009 SUPERVISED LEARNING We are given some training data: We must learn a function If y is discrete, we call it classification If it is

More information

Prediction of Hourly Solar Radiation in Amman-Jordan by Using Artificial Neural Networks

Prediction of Hourly Solar Radiation in Amman-Jordan by Using Artificial Neural Networks Int. J. of Thermal & Environmental Engineering Volume 14, No. 2 (2017) 103-108 Prediction of Hourly Solar Radiation in Amman-Jordan by Using Artificial Neural Networks M. A. Hamdan a*, E. Abdelhafez b

More information

Introduction Neural Networks - Architecture Network Training Small Example - ZIP Codes Summary. Neural Networks - I. Henrik I Christensen

Introduction Neural Networks - Architecture Network Training Small Example - ZIP Codes Summary. Neural Networks - I. Henrik I Christensen Neural Networks - I Henrik I Christensen Robotics & Intelligent Machines @ GT Georgia Institute of Technology, Atlanta, GA 30332-0280 hic@cc.gatech.edu Henrik I Christensen (RIM@GT) Neural Networks 1 /

More information

MODELLING TRAFFIC FLOW ON MOTORWAYS: A HYBRID MACROSCOPIC APPROACH

MODELLING TRAFFIC FLOW ON MOTORWAYS: A HYBRID MACROSCOPIC APPROACH Proceedings ITRN2013 5-6th September, FITZGERALD, MOUTARI, MARSHALL: Hybrid Aidan Fitzgerald MODELLING TRAFFIC FLOW ON MOTORWAYS: A HYBRID MACROSCOPIC APPROACH Centre for Statistical Science and Operational

More information

Machine Learning. Neural Networks. (slides from Domingos, Pardo, others)

Machine Learning. Neural Networks. (slides from Domingos, Pardo, others) Machine Learning Neural Networks (slides from Domingos, Pardo, others) For this week, Reading Chapter 4: Neural Networks (Mitchell, 1997) See Canvas For subsequent weeks: Scaling Learning Algorithms toward

More information

Weighted Fuzzy Time Series Model for Load Forecasting

Weighted Fuzzy Time Series Model for Load Forecasting NCITPA 25 Weighted Fuzzy Time Series Model for Load Forecasting Yao-Lin Huang * Department of Computer and Communication Engineering, De Lin Institute of Technology yaolinhuang@gmail.com * Abstract Electric

More information

From Last Meeting. Studied Fisher Linear Discrimination. - Mathematics. - Point Cloud view. - Likelihood view. - Toy examples

From Last Meeting. Studied Fisher Linear Discrimination. - Mathematics. - Point Cloud view. - Likelihood view. - Toy examples From Last Meeting Studied Fisher Linear Discrimination - Mathematics - Point Cloud view - Likelihood view - Toy eamples - Etensions (e.g. Principal Discriminant Analysis) Polynomial Embedding Aizerman,

More information

Machine Learning for OR & FE

Machine Learning for OR & FE Machine Learning for OR & FE Regression II: Regularization and Shrinkage Methods Martin Haugh Department of Industrial Engineering and Operations Research Columbia University Email: martin.b.haugh@gmail.com

More information

CONTEMPORARY ANALYTICAL ECOSYSTEM PATRICK HALL, SAS INSTITUTE

CONTEMPORARY ANALYTICAL ECOSYSTEM PATRICK HALL, SAS INSTITUTE CONTEMPORARY ANALYTICAL ECOSYSTEM PATRICK HALL, SAS INSTITUTE Copyright 2013, SAS Institute Inc. All rights reserved. Agenda (Optional) History Lesson 2015 Buzzwords Machine Learning for X Citizen Data

More information

Operations Management

Operations Management 3-1 Forecasting Operations Management William J. Stevenson 8 th edition 3-2 Forecasting CHAPTER 3 Forecasting McGraw-Hill/Irwin Operations Management, Eighth Edition, by William J. Stevenson Copyright

More information

Statistics Toolbox 6. Apply statistical algorithms and probability models

Statistics Toolbox 6. Apply statistical algorithms and probability models Statistics Toolbox 6 Apply statistical algorithms and probability models Statistics Toolbox provides engineers, scientists, researchers, financial analysts, and statisticians with a comprehensive set of

More information

APPLICATION OF RADIAL BASIS FUNCTION NEURAL NETWORK, TO ESTIMATE THE STATE OF HEALTH FOR LFP BATTERY

APPLICATION OF RADIAL BASIS FUNCTION NEURAL NETWORK, TO ESTIMATE THE STATE OF HEALTH FOR LFP BATTERY International Journal of Electrical and Electronics Engineering (IJEEE) ISSN(P): 2278-9944; ISSN(E): 2278-9952 Vol. 7, Issue 1, Dec - Jan 2018, 1-6 IASET APPLICATION OF RADIAL BASIS FUNCTION NEURAL NETWORK,

More information

Radial Basis Functions Networks to hybrid neuro-genetic RBFΝs in Financial Evaluation of Corporations

Radial Basis Functions Networks to hybrid neuro-genetic RBFΝs in Financial Evaluation of Corporations Radial Basis Functions Networks to hybrid neuro-genetic RBFΝs in Financial Evaluation of Corporations Loukeris Nikolaos University of Essex Email: nikosloukeris@gmail.com Abstract:- Financial management

More information

(Feed-Forward) Neural Networks Dr. Hajira Jabeen, Prof. Jens Lehmann

(Feed-Forward) Neural Networks Dr. Hajira Jabeen, Prof. Jens Lehmann (Feed-Forward) Neural Networks 2016-12-06 Dr. Hajira Jabeen, Prof. Jens Lehmann Outline In the previous lectures we have learned about tensors and factorization methods. RESCAL is a bilinear model for

More information

Advances in promotional modelling and analytics

Advances in promotional modelling and analytics Advances in promotional modelling and analytics High School of Economics St. Petersburg 25 May 2016 Nikolaos Kourentzes n.kourentzes@lancaster.ac.uk O u t l i n e 1. What is forecasting? 2. Forecasting,

More information

Module 12. Machine Learning. Version 2 CSE IIT, Kharagpur

Module 12. Machine Learning. Version 2 CSE IIT, Kharagpur Module 12 Machine Learning Lesson 39 Neural Networks - III 12.4.4 Multi-Layer Perceptrons In contrast to perceptrons, multilayer networks can learn not only multiple decision boundaries, but the boundaries

More information

Lecture 4: Feed Forward Neural Networks

Lecture 4: Feed Forward Neural Networks Lecture 4: Feed Forward Neural Networks Dr. Roman V Belavkin Middlesex University BIS4435 Biological neurons and the brain A Model of A Single Neuron Neurons as data-driven models Neural Networks Training

More information

Study of a neural network-based system for stability augmentation of an airplane

Study of a neural network-based system for stability augmentation of an airplane Study of a neural network-based system for stability augmentation of an airplane Author: Roger Isanta Navarro Annex 1 Introduction to Neural Networks and Adaptive Neuro-Fuzzy Inference Systems (ANFIS)

More information

Lecture 7 Artificial neural networks: Supervised learning

Lecture 7 Artificial neural networks: Supervised learning Lecture 7 Artificial neural networks: Supervised learning Introduction, or how the brain works The neuron as a simple computing element The perceptron Multilayer neural networks Accelerated learning in

More information

Prediction of Vegetable Price Based on Neural Network and Genetic Algorithm

Prediction of Vegetable Price Based on Neural Network and Genetic Algorithm Prediction of Vegetable Price Based on Neural Network and Genetic Algorithm Changshou uo 1, Qingfeng Wei 1, iying Zhou 2,, Junfeng Zhang 1, and ufen un 1 1 Institute of Information on cience and Technology

More information

Journal of of Computer Applications Research Research and Development and Development (JCARD), ISSN (Print), ISSN

Journal of of Computer Applications Research Research and Development and Development (JCARD), ISSN (Print), ISSN JCARD Journal of of Computer Applications Research Research and Development and Development (JCARD), ISSN 2248-9304(Print), ISSN 2248-9312 (JCARD),(Online) ISSN 2248-9304(Print), Volume 1, Number ISSN

More information

Multilayer Neural Networks. (sometimes called Multilayer Perceptrons or MLPs)

Multilayer Neural Networks. (sometimes called Multilayer Perceptrons or MLPs) Multilayer Neural Networks (sometimes called Multilayer Perceptrons or MLPs) Linear separability Hyperplane In 2D: w x + w 2 x 2 + w 0 = 0 Feature x 2 = w w 2 x w 0 w 2 Feature 2 A perceptron can separate

More information

Ch.6 Deep Feedforward Networks (2/3)

Ch.6 Deep Feedforward Networks (2/3) Ch.6 Deep Feedforward Networks (2/3) 16. 10. 17. (Mon.) System Software Lab., Dept. of Mechanical & Information Eng. Woonggy Kim 1 Contents 6.3. Hidden Units 6.3.1. Rectified Linear Units and Their Generalizations

More information

Efficient Forecasting of Exchange rates with Recurrent FLANN

Efficient Forecasting of Exchange rates with Recurrent FLANN IOSR Journal of Computer Engineering (IOSR-JCE) e-issn: 2278-0661, p- ISSN: 2278-8727Volume 13, Issue 6 (Jul. - Aug. 2013), PP 21-28 Efficient Forecasting of Exchange rates with Recurrent FLANN 1 Ait Kumar

More information