The Combination of Traditional Asset Pricing Models and Machine Learning

This article combines the CAPM theory with the concept of beta coefficient, proposes a multiple linear regression model and recurrent neural network (RNN), and predicts the stock of representative Apple company in SP500. The multiple linear models introduce the concept of CAPM (beta) and is based on multiple linear regression. It determines the values of these coefficients by minimizing the error between actual and predicted values. The LinearRegression class in the sklearn library is used to train and predict data related to Apple Inc. Recurrent neural network (RNN) is used to predict the stock price of Apple Inc. and combines the beta coefficient calculated from market index data to enhance the performance of the model. The results show that the predicted values are very close to the actual values. In addition, this article also compared and demonstrated the different prediction results of multiple linear regression models and recurrent neural networks (RNN) on whether to introduce CAPM related concepts (beta). The results showed that the citation of related CAPM concepts is very necessary, and it is particularly strong in recurrent neural network (RNN) models. Afterwards, this article diverged from this result and further demonstrated from both positive and negative perspectives whether more parameters would bring better predictive results to the model.