A TOPSIS approach of ranking classifiers for stock index price movement prediction

Abstract

Predicting future stock index price movement is equivalent to a binary classification problem with one class label for increasing movement and other for decreasing movement. In the literature, a wide range of classifiers are tested for this application, but the decision regarding a better technique varies with the choice of performance measures. Hence, assessing classifiers can be considered as a multi-criteria decision-making (MCDM) problem. In this study, a TOPSIS-based MCDM framework is suggested for ranking five classifier models such as radial basis function, Naïve Bayes, decision tree, support vector machine, and k-nearest neighbor with respect to four criteria in application to prediction of future stock index price movements. Historical stock index prices of two benchmark stock indices such as BSE SENSEX and S&P500 are taken for the empirical validation of the model. The results reveal that ranking a classifier with respect to multiple evaluation measures is better compared to selecting one considering single criterion.