Empirical study on the performance of various machine learning models in predicting stock price movements as a binary classification task

Abstract

This paper examines the accuracy of stock price rise-or-fall predictions of seven different machine learning algorithms, including support vector machines and random forests, for three industry types: securities, banks, and Internet companies. The purpose of the research is to explore the effects of different models in the stock market, so as to help people choose the optimal machine learning model in predicting different types of stocks. The study produced nine features based on the study by Patel et al for prediction. By collecting 9 types of stock data from companies in different industries, we performed necessary preprocessing on the data, fitted the model, tuned the parameters of the model and get the prediction result. Through the result, we found that the random forest algorithm has obvious advantages in binary classification prediction of stock prices. Linear discriminant analysis (LDA), Quadratic Discriminant Analysis (QDA) and logistic regression also have good fitting effects in this type of problem. K-Nearest Neighbor (KNN) and Naive Bayes algorithms exhibit poor prediction accuracy.