基于Attention-GRU的量化选股策略研究
Research on Quantitative Stock Selection Strategy Based on Attention-GRU

本文选取上证50成分股作为股票池，根据对应时间段内相关的因子和股票数据，运用基于注意力机制的门控循环单元神经网络模型进行分类预测，建立多因子选股模型，构建选股策略，最终对构建的投资组合进行回测。策略兼顾了风险和收益，在控制风险的同时，能够获得超额回报。
In this paper, 50 constituent stocks of Shanghai Stock Exchange are selected as the stock pool. According to the relevant factors and stock data in the corresponding time period, the gated recurrent unit network model based on attention model is used for classification prediction, the multi-factor stock selection model is established, the stock selection strategy is constructed, and finally the constructed portfolio is back tested. The strategy takes into account the risk and return, and can obtain excess return while controlling the risk.