Time Series Forecasting: A Comparative Study of VAR ANN and SVM Models

摘要

Modeling and forecasting time series data has primarysignificance to numerous practical areas. Numerous significant models have beenproposed in texts for enlightening the correctness and effectiveness ofmodeling and predicting time series data. The goal of this work is to discoveryan appropriate model to forecast time series data. Firstly, we applied the mostpopular multivariate time series model is Vector Autoregressive model, with itsfrequently used four criteria of VAR order selection such as AIC, HQ, SC andFPE,? and the asymptotic Portmanteau testof VAR order selection. We also checked the accuracy of forecast performancebased on impulse response function. Secondly, we applied the most popular timeseries modeling and forecasting machine learning techniques, such as artificialneural network and support vector machine. In this study, we applied the threemultivariate time series models, viz. VAR, ANN and SVM, compared together withtheir inherent forecasting strengths based on the five forecast performancemeasures: mean squared error, mean absolute deviation, root mean squared error,mean absolute percentage error and Theil’s U-statistics. Finally, we found thatthe artificial neural network, time series modeling and forecasting machinelearning technique, is the best technique for time series modeling andforecasting.