Stock Price Forecasting Method based on Features from the Relevant Factors

Predicting stock trend by using machining learning is a hot research issue today. However, it is still an open problem which features, or which combination of features is the best for the trend predication of price. In this research, a method of stock price prediction based on stock information is proposed, we choose 249 features from the stock of Baolihua New Energy ranging from 2019 to 2020. Five methods are used to convolute and assess those features, and then top 30 correlative features are sent to support vector regression for training. The back test, on the 2021 stock data, indicates that annual yield of comprehensive method reaches 44.4%, while yield of simplex methods are only approximately 20%, based on the benchmark yield of −9%. The result shows that this stock price forecasting method deserves further investigation.