Forecasting stock price movement direction by machine learning algorithm

<p><span lang="EN-US">Forecasting stock price movement direction (SPMD) is an essential issue for short-term investors and a hot topic researchers. It real challenge concerning the efficient market hypothesis that historical data would not be helpful in forecasting because it already reflected prices. Some commonly-used classical methods are based on statistics econometric models. However, becomes more complicated when variables model all nonstationary, relationships between sometimes very weak or simultaneous. The continuous development of powerful algorithms features machine learning artificial intelligence has opened promising new direction. This study compares predictive ability three models, including <a name="_Hlk106797328"></a>support vector (SVM), neural networks (ANN), logistic regression. used those stocks VN30 basket with holding period one day. With rolling window method, this got highly SVM average accuracy 92.48%.</span></p>