Multi-factor Investment Model Based on TabNet

Abstract

Abstract The prediction of the stock market is a complex and nonlinear prediction, which is affected by multiple information sources. However, the traditional investment method can only rely on people’s subjective experience to invest and select stocks. In this way, quantitative investment shows itself in various fields with the huge quantitative operation ability of computers. In the past research on quantitative investment, the methods of decision trees or neural networks have their own shortcomings. In this paper, TabNet is used to simulate the process of decision tree in the way of multi-factor investment. We select the data of China Securities Index (CSI) 300 as the data set. Through this method, we improve the problems of low precision of decision tree and easy overfitting of neural network. The empirical results show that the model has good prediction performance, which is better than the currently used LSTM and Gan and Gated Recurrent Units(GRU) models, and the model also has good interpretability.