Stock Selection And Timing-choose Portfolio Investment Strategy Based On K-Means Volatility Clustering

Quantitative Quantitative trading has become increasingly mature abroad and developed rapidly in China.It has become a hot topic in recent years.At present,many domestic fund companies spend a lot of energy on quantitative research to build stable quantitative strategies.Quantitative trading is divided into two parts:quantitative stock selection and quantitative timing.The combination of these two parts can reduce the risk of quantitative strategy and stabilize and improve the return of quantitative strategy.In this paper,the K-Means algorithm in machine learning is used to cluster and mark the stock volatility.The marked stocks of each class are used as the stock pool of the quantitative timing strategy,and combined with the original quantitative timing strategy,and then a complete stock timing portfolio investment strategy is constructed.The back-test tool is used to back-test the original timing strategy and the combined portfolio investment strategy,Then compare and evaluate the two strategies through the quantitative strategy evaluation index,so as to prove the effectiveness of K-Means clustering stock volatility as a quantitative stock selection to improve the original timing strategy,and finally build a stable quantitative investment strategy that can be applied to firm trading.In addition,quantitative trading is a process of constant adjustment.The sample research data is too far away from the time of real trading,and the strategy will lose its effectiveness.Therefore,the back-test time of this paper is selected from January 11,2017 to January 11,2022.The strategy formed in this way is more in line with the situation of China’s stock market and closer to reality.Volatility measurement adopts three types of measurement methods: 5ATR,GK standard volatility and realized volatility(RV)to measure daily volatility.The original strategy selected the relatively simple MACD strategy,Brinband strategy and the RSRS timing strategy developed by Everbright Securities.The CSI 300,CSI 500 index and GEM index were used as the benchmark,and its constituent stocks were used as the stock pool of the original strategy.The three original strategies were tested back and the results were recorded.Measure the daily volatility of the constituent stocks of the Shanghai and Shenzhen 300,the China Securities 500 Index and the GEM index using the above three volatility measurement methods,add the volatility data of the constituent stocks of each index in the sample interval to the K-Means cluster model,calculate the evaluation index SSE and contour coefficient,and make the curve chart of the evaluation index,and select the best K of each index constituent stock through the elbow method and the contour coefficient.Take the optimal K selected in the sample set as the K value of the K-Means algorithm in the training set,put the daily data of the index component stocks into the K-Means model for clustering,and then rank each category according to the size of the cluster center from low to high,and finally mark the component stocks in each category.During the back test,the stocks of different volatility measurement methods and different K-Means categories are used as the stock pools of three original strategies respectively,The original strategy is combined and compared under the condition that other back-test parameters remain unchanged.According to the comparison of the back-test results of the stock selection and timing portfolio investment strategy and the original strategy,it is found that the timing strategy after stock selection based on K-Means volatility has great changes in both the return and risk.In different target and different timing strategies,the volatility information required for the timing strategy to make judgments is different,and different volatility measurement methods can reflect the volatility information of each stock between the plates,According to the style of quantitative strategy,we should use more volatility measurement methods to match the style and target of quantitative timing strategy as much as possible,and find the best one.The ATR volatility measurement method adopted in this paper works better after combining with timing strategy among stocks with higher volatility level,and the RV and GK standard volatility works better after combining with timing strategy among stocks with lower volatility level.The strategy improved by K-Means volatility clustering is actually a division of the original strategy.All categories of K-Means volatility clustering improvement strategies are combined to restore the original strategy,which logically classifies strategies with different risk levels and different income levels,Investors or quantitative institutions can choose the category with the highest yield as the stock pool of quantitative strategy according to the risk they can bear.This category of strategy is the best for them.Finally,the effect of K-Means volatility clustering method in improving the original timing strategy is different,and the effect of different timing strategies is also different.Institutional investors should adjust according to the specific situation of the strategy when using this method...