Testing Stock Market Efficiency Using Historical Trading Data and Machine Learning

Abstract: Stock forecasting is a problem that is important in finance because it aids investors in financial decision making. According to the efficient market hypothesis stock markets are efficient in such a way that it's impossible to gain excess returns over the market by making decisions based on current available information. This paper evaluates the usage of machine learning algorithms and historical trading data for stock price prediction combined with investment strategies in order to test the efficient market hypothesis. The results show that none of the tested machine learning algorithms managed to gain excess returns over the market which confirms the efficient market hypothesis.