Empirical Research Of Stock Selection Strategy Based On Machine Learning

With the development of economic globalization,China's capital market has become increasingly complex,its scale has gradually expanded in the process of development,various types of investment in the market emerge in endlessly,the number of investment has become increasingly large,regardless of whether the source of investment funds is individuals or institutions,the investment they own.Capital channels and ways are also on the rise.Among them,a new way of investment-Quantitative investment,is attracting more and more people's attention.Quantitative investment originates from foreign capital market,and has caused a great wave in overseas capital market with a brand-new advanced technology.This paper mainly studies how to combine artificial intelligence with the traditional multi-factor model to construct a quantitative investment strategy based on artificial intelligence stock selection.Through this strategy,the investment portfolio is constructed by selecting the stocks with investment value from the listed stocks,so as to obtain stable high return in this way.This strategy has obvious advantages,which can not only expand the content of the traditional model of factor selection,but also give direction and data when researchers explore deeply.Starting from this aspect,this paper chooses the factor cross-section data of trading days from January 2015 to July 2018 as the data sample.The construction process is mainly divided into data pre-processing and construction model to verify the feasibility of model construction.The results obtained by using this model are analyzed and optimized comprehensively.In this paper,linear regression,ridge regression,SVR and random forest are used to construct the strategy.The quantitative investment strategy based on AI stock selection model constructed in this paper between January 2014 and July 2018,the best yield is close to 700%,annual return is also close to 60%,the performance of the same period of comparison of the Chinese stock index is far behind.Through grouping comparative analysis,we can see that grouping changes will have a profound impact on strategic performance,and there is a negative correlation between the two.It further shows that the model studied in this paper has a significant effect on the item of stock classification,and can perfectly identify the differences between the dominant and the disadvantaged stocks.By comparing and analyzing the strategies under different factors,different positions and different market styles,it can be found that the performance of the quantitative investment strategy based on artificial intelligence is quite different.In addition,in the aspect of strategy optimization,the best performance optimization strategy is selected through parameter optimization and training set length modification.