Research On Multi-factor Stock Selection Strategy Based On ESG Factor Fusion Mode

Environmental,Social and Governance(ESG)is a sustainable development value that emphasizes the coordinated development of various aspects of economy,environment,and society.This is highly consistent with a series of brilliant ideas,great visions,and grand goals proposed at the 20 th National Congress of the Communist Party of China,which has promoted the transformation of enterprises from pursuing the maximization of shareholder interests to pursuing the maximization of social value.Therefore,ESG investment has flourished in China,gradually becoming the mainstream investment philosophy.Therefore,in today’s China,where the capital market system is gradually improving,excellent enterprises are blooming,and quantitative investment is rising rapidly,how can investors rationally choose high-quality stocks that actively assume social responsibility to invest in the financial market in accordance with the development of the times,and thus achieve relatively stable long-term returns? This is a question worthy of in-depth study.Based on the multi factor stock selection model,this paper optimizes its two key steps: factor selection and model fitting.In factor selection,ESG factors are added to build candidate factor pools based on traditional factors;In model fitting,a two-level Stacking fusion model is built.The first layer uses five single machine learning algorithms,namely,decision tree,random forest,gradient lifting decision tree,XGBoost,and Light GBM,as the base model.The second layer uses logical regression as the metamodel,with the output of the base model as the input feature of the metamodel,and the output of the metamodel as the final result.Based on this,a total of four multi factor stock selection strategies are constructed for the reference group and the optimization group.By comparing the differences between different strategies,it is explored whether introducing ESG factors into the multi factor stock selection model can screen high-quality stocks that match the background of the times,and whether the fitting process of optimizing the multi factor stock selection model using Stacking integrated learning algorithm can better improve the prediction accuracy and stock selection ability of the model than a single machine learning algorithm.Through empirical research,conclusions are drawn: First,ESG factors can be used as effective factors in multi factor stock selection models;Secondly,introducing ESG factors on the basis of traditional factors in the multi factor stock selection model can bring its own stock selection advantages into play and obtain excess returns;Thirdly,the multi factor stock selection strategy constructed based on the Stacking ensemble learning algorithm is superior to the multi factor stock selection strategy constructed by each single machine learning algorithm in terms of model prediction accuracy and stock selection ability.In view of this,this article provides relevant suggestions from the regulatory level,enterprise level,and investment level.