Asset Price Prediction via Machine-Learning Method: A Review

Abstract

This paper aims at a brief introduction of application and research progress of machine learning in asset pricing. Firstly, we start with the two well-known classical models of asset pricing, CAPM model and APT model, which describe relationship between the expected return of effective portfolio and risk assets in the capital market. That is the basis of asset pricing theory. Then, the key point locates on summarizing the applicability and limitations of different machine-learning models. Thirdly, by using out-of-sample R2 to measure the performance of each model, we examine how machine learning can be used to predict the risk premium. At the same time, it is demonstrated that the two best-performance methods with nonlinear predictor interactions are random forests and neural networks. Finally, we suggest the future research trends for the application of machine learning method in asset price prediction.