一种基于机器学习和蜡烛图的股市投资策略 研究

There is no consistent conclusion whether the candlestick pattern can predict the market trend effectively. This paper discusses 8 common two-day candlestick patterns to make long position predictions in Chinese stock market, and then 5 common machine learning methods including LR, KNN, RF, SVM and RBM are applied to enhance the predictive power of the model. Based on this method, a short-term investment strategy is constructed. Empirical results show that six of these candlestick patterns have predictive power of long position in Chinese stock market. Accuracy of five patterns has been enhanced to varying degrees and produces better stability after the processing of machine learning methods. In addition, we conclude that investment returns would be reduced if trading frequently.