A stock multi-factor model based on slice inverse regression and the Bootstrap method

Abstract

With the increasingly prominent position of machine learning algorithms in the field of financial quantification, the research on factor quantification and prediction models for high-frequency stock price time series has emerged as a prominent area of study. This paper presents a novel factor quantization prediction model based on slice inverse regression and Bootstrap. The proposed method effectively addresses the challenge of data dimensionality disaster and enables factor quantification while preserving the information of the original predictor variables. Additionally, the method incorporates Bootstrap technology and adopts the concept of model averaging instead of model selection, which enables the adaptive capturing of the unknown connection structure between factors and stock prices. Moreover, it effectively balances the bias and variance of individual prediction models. The empirical analysis using actual data demonstrates the superiority of the proposed method over the comparison method in terms of evaluation indicators such as mean square error, thus highlighting its robustness.