Construction of Investment Strategies for WIG20, DAX and Stoxx600 with Random Forest Algorithm

Abstract

Machine learning provides powerful tools for data analysis, especially in regression and classification problems what may be used in creation of investment strategies. This paper present an efficient way of utilization of one of the machine learning algorithms on examples of stock indices: Stoxx600, WIG20 and DAX. This work concentrates on time series analysis of stock indices with Random Forest algorithm to create investment strategies based on future probabilities of declines and upswings. Taking into account some macroeconomic characteristics, technical indicators and consensus estimates, the models are trained to provide a buy signal if the output probability is above a specific threshold and sell signal in case of the opposite situation. The examination of the strategies efficiency indicates the differences in determinants among chosen stock indices.