基于Elastic Net分位数回归的多因子量化选股策略

Multi-Factor Strategy Based on Elastic Net Quantile Regression

将Elastic Net分位数回归应用到多因子选股中,求解过程采用SNCD算法。选取2013年6月28日至2021年7月1日的沪深300指数成分股,选取了共46个因子,回测结果表明,Elastic Net分位数回归在分位点为0.1和0.9时,年化收益率分别达到了38.51%和39.67%,远超基准年化收益率17.37%。同时还将Elastic Net分位数回归策略同Lasso分位数回归策略比较,从回测结果的各项指标来看,Elastic Net分位数回归策略可以通过调整不同的分位点来保留更加有效的因子,从而获得更高的年化收益率和超额收益率。

Elastic net quantile regression is applied to multi factor stock selection, and the solution process adopts sncd algorithm. A total of 46 factors were selected from the constituent stocks of the Shanghai and Shenzhen 300 index from June 28, 2013 to July 1, 2021. The back test results showed that when the quantile regression of elastic net was 0.1 and 0.9, the annualized yield reached 38.51% and 39.67%respectively, far exceeding the benchmark annualized yield of 17.37%. At the same time, the elastic net quantile regression strategy is also compared with lasso quantile regression strategy. From the indicators of the back-test results, the elastic net quantile regression strategy can retain more effective factors by adjusting different quantiles, so as to obtain higher annualized return and excess return.