国证 2000 指数增强策略 经过因子测试与筛选，包括技术、反转、特异波动率等在内的因子在国证 2000 指数成分股上均有出色表现，我们所 合成的各个大类因子也基本都起到了很好的提升效果。10 月该因子表现恢复出色，IC 值 25.34%。样本外整体策略表 现出色，10 月策略的超额收益为 2.92%。 基于多目标、多模型的机器学习指数增强策略 根据国金金融工程团队发布的《基于多目标、多模型的机器学习指数增强策略》，原策略中我们选取了 GBDT 和 NN 两 大类结构具有一定差异的模型，选取不同的特征数据集进行分别训练，并使用多种预测标签进行对比并融合，最终 构建出的 GBDT+NN 机器学习选股因子在 A 股各类宽基指数上历史表现优异。但在今年以来，尤其是近期市场风格出 现调整后有失效表现。 对此，我们根据《Alpha 掘金系列之十八：基于 TimeMixer 改进的选股因子到 ETF 轮动策略》，创新性地将其多尺度 混合与季节/趋势分解机制引入 GRU 模型，通过 LightGBM 集成 TSGRU 隐向量与传统量化因子，构建了改进的机器学习 选股模型，该模型能更好地捕捉近期的市场信息，表现出色。 风险提 ...

- The recent phenomenon of "strong index, weak quantitative Alpha" is attributed to style mismatches, with cumulative excess returns driven by small-cap and short-term momentum factors initially, and later by analyst consensus expectations and growth styles[2][3] - The Guozheng 2000 Index enhancement strategy involves factor testing and selection, including technical, reversal, and idiosyncratic volatility factors, which have shown excellent performance in the Guozheng 2000 Index constituents[4] - The machine learning index enhancement strategy based on multiple objectives and models uses GBDT and NN models, trained on different feature datasets and combined to construct a GBDT+NN stock selection factor, which has performed well across various broad-based indices in the A-share market[5] - The dividend style timing + dividend stock selection fixed income+ strategy uses 10 indicators related to economic growth and monetary liquidity to construct a dynamic event factor system for dividend index timing, showing significant stability improvement compared to the CSI Dividend Index total return[6] - The Guozheng 2000 Index enhancement factor's IC mean is 12.54%, with a T-statistic of 12.56, indicating good predictive performance[4] - The GBDT+NN machine learning stock selection factor in the CSI 300 constituents has an IC mean of 11.43% and an annualized excess return of 15.39%[43] - The GBDT+NN machine learning stock selection factor in the CSI 500 constituents has an IC mean of 9.77% and an annualized excess return of 29.48%[48] - The GBDT+NN machine learning stock selection factor in the CSI 1000 constituents has an IC mean of 13.49% and an annualized excess return of 16.10%[53] - The Guozheng 2000 Index enhancement strategy has an annualized excess return of 13.18% and an IR of 1.73[38] - The GBDT+NN CSI 300 Index enhancement strategy has an annualized excess return of 10.86% and an IR of 1.81[47] - The GBDT+NN CSI 500 Index enhancement strategy has an annualized excess return of 10.27% and an IR of 1.71[52] - The GBDT+NN CSI 1000 Index enhancement strategy has an annualized excess return of 15.83% and an IR of 2.34[57] - The dividend stock selection strategy has an annualized return of 18.83% and a Sharpe ratio of 0.89[58] - The dividend timing strategy has an annualized return of 13.58% and a Sharpe ratio of 0.88[58] - The fixed income+ strategy has an annualized return of 7.34% and a Sharpe ratio of 2.17[58]