Assessing efficacy of association rules for predicting global stock indices

Abstract

Present study explores the efficacy/performance of association rules for prediction of global stock indices. Global stock indices data for the last 12 years are used to develop the prediction models. The data consists of several technical indicators. Technical indicators were converted to categorical variables and rules were extracted using association rules. The performance of mined rules was tested for global stock indices considered in this study. Based on the findings of the study, it can be concluded that association rules have potential to provide profitable returns with a fair degree of model parsimony. The outcome of the study indicate that Stochastic Oscillator %K%D, relative strength index (RSI), Disparity 5 Days and Disparity 10 Days are the common market signal sources across all stock indices. Along with these, investors can make decisions using additional indications from rate of change (ROC), commodity channel index (CCI) and Momentum. Association rules can be used for profitable decision making with limited number of technical indicators. Limited number of technical indicators are easy to handle even for smaller retail investors. Trading decisions made on the basis of mined association rule were able to comprehensively beat buy-and-hold return for the selected indices included in the study.