Research On Multi-factor Quantitative Stock Selection Strategy Based On Dynamic Contextual Model

With the continuous opening of China’s capital market and the awakening of people’s financial management awareness,more and more domestic investors have begun to pay attention to quantitative investment,an investment method widely used in Western markets.Multi-factor stock selection is one of the most important methods of quantitative investment.Investors study the financial market,dig out factors related to stock returns and construct investment models to select stocks that are expected to perform well in order to obtain returns that exceed the benchmark.Traditional multi-factor stock selection is often a unified evaluation of all stocks in the stock pool,ignoring the applicability of factors in different styles of stock pools.At the same time,the fundamentals of individual stocks are very different and constantly changing,and a one-size-fits-all evaluation of stocks is also not conducive to accurate capture the characteristics of individual stocks.In order to solve this problem,Sorensen,Hua,and Qian took a different approach.Instead of selecting factors,they proposed a dynamic contextual theory,which divided stocks into different contexts based on fundamental and technical information.In different contexts,the same factors are of different importance to forecast earnings.This method can better describe the characteristics of stocks and select stocks that have a high probability of exceeding the benchmark.This article further designs a multi-factor stock selection model with reference to the dynamic contextual theory.From the perspective of fund companies,this article will study the stock returns and factor from January 2016 to December 2019 according to the different trends of China’s A-share market during the sample period.First,the relevant factors of the tradable stocks in China’s A-share market are divided into contexts such as quality,return and risk,value,sentiment,technical indicators,per share indicators,growth rate and other indicators.Individual stocks are classified as "high" context or "low" context.In each context,the weight of the factor is different,and the weight of the factor will be adjusted every month based on the performance of the previous month.Through calculation,5 contexts division factors and 8 Alpha factors are obtained.Taking the CSI 300 constituent stocks as the object,relying on the UQER platform to establish a multi-factor quantitative stock selection model based on the dynamic contextual model,using the period from January 2020 to January 2021 to back-tested.The stocks are selected by calculating the scores of the stocks in the model,and the adaptability of the model is tested according to indicators such as the actual investment return and information ratio of the portfolio.In addition,this paper compares the multi-factor equal-weight model and the dynamic contextual multi-factor model to further verify the feasibility and effectiveness of the strategy.Studies have shown that the stock portfolio obtained by the dynamic contextual multi-factor quantitative stock selection model can outperform the benchmark with a high probability and achieve better returns.