A Novel Interpretable Stock Selection Algorithm for Quantitative Trading

In recent years, machine learning models have exhibited remarkable performance in the fourth industrial revolution. However, especially in the field of stock forecasting, most of the existing models demonstrate either relatively weak interpretability or unsatisfactory performance. This paper proposes an interpretable stock selection algorithm (ISSA) to achieve accurate prediction results and high interpretability for stock selection. The excellent performance of ISSA lies in its integration of the learning to rank algorithm LambdaMART with the SHapley Additive exPlanations (SHAP) interpretation method. Performance evaluation over the Shanghai Stock Exchange A-share market shows that ISSA outperforms regression and classification models in stock selection performance. The results also demonstrate that the proposed ISSA solution can effectively filter out the most impactful features, potentially used for investment strategy.