Analysis Of Multi Factor Stock Selection Strategy Based On Random Forest Method

With the development of China's economy,the complexity of China's capital market continues to deepen,and the scale of the investment market is growing.For investors,whether individual investors or institutional investors,the types,channels and ways of investment are also growing day by day.In this context,investors will inevitably choose more advanced investment methods to obtain excess return.At the same time,with the rapid development of the computer industry and the advent of the big data era,quantitative investment,which is a new thing produced by the combination of Finance and computer,has become one of the most popular investment methods in the financial industry.In this thesis,select 300 stock pools of Shanghai and Shenzhen as the basic data,and increase the types and numbers of factors to a certain extent on the basis of predecessors.We select a total of 24 factors from the five categories of valuation,growth,financial quality,technology and leverage to build a factor pool,and then build a multi factor stock selection model through random forest algorithm,and finally carry out strategic back testing analysis.Factor data is selected from January 2010 to December 2016.Factor data and stock return are extracted as sample data on the last trading day of each month.Among them,from January 2010 to December 2012,factor effectiveness analysis was conducted to ensure the effectiveness of the selected factors;from January 2013 to December 2014,traditional multi factor stock selection model was organically combined with machine learning algorithm to model the sample data using random forest algorithm and find the optimal parameters of the algorithm,so as to determine the optimal characteristic number and resolution of random forest algorithm The number of policy trees;from January 2015 to December 2016,the sample data were tested and compared,and the stock selection strategy based on the random forest method was compared with the traditional multi factor stock selection strategy to verify the effectiveness and practicability of the model,thus highlighting the advantages of the random forest algorithm in fault tolerance and avoiding over fitting.The multi factor stock selection strategy based on the random forest method constructed in this paper has achieved an annual yield of 118.11%,which is far higher than the yield of the market benchmark index(CSI 300)and the traditional multi factor stock selection strategy,and the beta value of the strategy is 0.91,which indicates that the strategy will not produce serious deviation under the fluctuation of system risk.On the whole,the strategy has the ability of stock selection,which can provide investors and investment institutions with higher excess return under unit risk,and provide new ideas and ideas for the design and development of stock selection strategy in the future.