Entropy-Based Technical Analysis Indicators Selection for International Stock Markets Fluctuations Prediction Using Support Vector Machines

Most of works on stock price forecasting are concerned with the problem of predicting its future value. However, forecasting stock price future fluctuation trend could be easier and interesting for traders and investors to maximize profits. The purpose of this study is to predict CAC40, FTSE, NASDAQ and S&P500 price index up and down fluctuations. In particular, it aims to propose a methodology to forecast regime switches in these markets time series to assist traders and investors in decision making. In the first stage, a large set composed of twenty five technical analysis indicators is formed. They fall into four broad categories namely oscillators, stochastic measures, indexes and indicators. Entropy statistic is employed to rank the initial technical analysis indicators. Finally, in the third stage, polynomial-based kernel support vector machines (SVM) are used for predicting CAC40, FTSE, NASDAQ and S&P500 future upward and downward fluctuations. The forecasting results show that the choice of technical analysis indicators used to predict CAC40 and NASDAQ fluctuations depend on the type of risk-aversion and risk-appetite of the investor. For the S&P500 and FTSE, technical analysis indicators used in our study can detect future downshifts with high accuracy. Thus, they are suitable for market analysis and trading by risk-averse investors on these markets.