Enhancing forecasting performance of multivariate time series using new hybrid feature selection

Abstract

The aim of this study is to propose a new hybrid feature selection model to improve the performance of multivariate time series (MTS) forecasting under uncertainty situation. This new hybrid model is called cooperative feature selection (CFS) and consists of two different component; GRA Analyzer and ANN Optimizer. The performance of CFS is evaluated on KLSE close price. The statistical analysis of the results shows that CFS has high ability to recognize and remove irrelevant input for obtaining optimum input factors, shortening the learning time and improving forecasting accuracy for vague MTS.