Using autoregressive modelling and machine learning for stock market prediction and trading

Abstract

Investors raise profit from stock market by maximising gains and minimising loses. The profit is difficult to raise because of the volatile nature of stock market prices. Predictive modelling allows investors to make informed decisions. In this paper, we compare four forecasting models: autoregressive integrated moving average (ARIMA), vector autoregression (VAR), long short-term memory (LSTM) and nonlinear autoregressive Exogenous (NARX). The results of predictive modelling are analysed and compared in terms of prediction accuracy. The research aims to develop a new profitable trading strategy. Our findings are: (i) the NARX model has provided accurate short-term predictions but failed long forecasts, and (ii) the VAR model can form a good trend line required for trading. Thus, the profitable trading strategy can combine the machine learning predictive modelling and technical analysis.