Application of fuzzy sets in time series forecasting

Abstract

In the modern competitive world, government and business organizations have to make the right decision in time depending on the information at hand. As large amounts of historical data are readily available, the need of performing accurate forecasting of future behavior becomes crucial to arrive at good decisions. Therefore, demand for a definition of robust and efficient forecasting techniques is increasing day by day. A successful time series forecasting depends on an appropriate model fitting. Time series data are highly non-stationary and uncertain in nature. Therefore, forecasting of time series using statistical or mathematical techniques is extremely difficult. The scientific community has been attracted by soft computing (SC) techniques in recent years to overcome these limitations. The SC is an amalgamation of different methodologies, such as fuzzy sets, neural computing, rough sets, evolutionary computing and probabilistic computing, to solve real world problems. The present work is a comprehensive examination of designing models for time series forecasting based on SC techniques, especially fuzzy time series (FTS). In this thesis, we provide in-depth study of various issues and problems associated with the FTS modeling approaches in time series forecasting. Apart from exhaustive literature survey on applications of FTS in time series forecasting, we provide improved methods for forecasting based on the FTS.