Comparison of Price Prediction Based on LSTM, GRU, Random Forest, LSSVM and Linear Regression

Abstract

Investment decision-making involves numerous factors to yield significant profit. Contemporarily, various models are proposed to be used in stock price prediction. However, the traditional linear model lacks the ability to mine the implicit information of the data, resulting in difficulties in deliver satisfactory performance on nonlinear data with large fluctuations and strong noises. LSTM, GRU, Optimized Random Forest, and LSSVR were employed as training methodologies to study the effectiveness in predicting directional movements of close stock prices of TESLA from July 2018 till July 2022 in comparison with the Linear regression model. This study adopted a combination of technical and fundamental analysis to reflect various sources of influence factors in the movement of stock prices. According to the analysis, the proposed models demonstrate a better accuracy score and excelled in avoiding a severe overfitting issue found in the benchmark algorism. These results shed light on guiding further exploration on machine learning techniques.