Stock price prediction based on multifactorial linear models and machine learning approaches

Abstract

With the development of economic and algorithmic models, stock price prediction has attracted more and more attention. This paper will investigate the stock forecasting based on multiple factors linear models and machine learning scenarios including four methods: multiple linear regression, exponential weighted moving average (EWMA), extreme gradient advance (XGBoost) and long-term short-term memory (LSTM). Three listed companies of Chinese market are selected from each of the following industries (total of nine companies) information technology, banking finance and catering and hotel. Besides, 730 daily market data is retrieved from January 1, 2019 to December 31, 2021. By using 18 factors with both daily market factors and technical factors to jointly predict the closing prices of these 9 stocks, the prediction effect of multiple linear regression model is better than XGBoost and LSTM in the case of less data. In other words, XGBoost and LSTM cannot give full play to their own advantages under this background. The research of this paper further explores the applicable scenarios of different models and these results shed light on the related research of multi model prediction of stock closing price.