Forecasting the Direction of Stock Market Index Movements Using Machine Learning Algorithms

Huang Jifei

Abstract

In financial sectors, forecasting the index movement of stock markets has always been regarded as a challenging task. Hence, there is a great deal of research being conducted regarding the prediction of the direction of stock price index movement. An increasing number of researchers are trying to adopt different machine learning algorithms to achieve a higher hit ratio prediction. In this project, we are going to use the same predictors and methodology in one research paper titled “Forecasting stock market movement direction with support vector machine” by Huang with his team as a base line model, and we are going to replicate the results for NIKKEI 225 index. Next, based on the existing results (from the base line model), we aim to make further improvements. To implement the further improvements, a total of 3 models have been built: (1) KNN and (2) Naïve Bayes models and (3) a combined model consisting of SVM, KNN and Naïve Bayes (NB). From the results, we observed that KNN performed better than Naïve Bayes whereas the combined model achieved the best hit ratio (72%). Next, in order to test the robustness of using machine learning to predict the stock index movement direction, we apply the model on another well-established New York stock market. We first select a set of macroeconomic predictors (e.g. inflation, unemployment rate, real GDP, private consumption), then conduct a pre-processing process (e.g. PCA and correlation matrix), and finally fit it into the classification model. Based on the prediction results, the combined model of SVM, KNN and NB has a strong predicting capability for the direction of stock market index movement.