Quantitative Models and Construction - **Model Name**: DecompGRU **Model Construction Idea**: The model improves information interaction between time-series and cross-sectional data by introducing two simple de-mean modules on the GRU baseline model[17] **Model Construction Process**: 1. The DecompGRU model is based on the GRU baseline architecture 2. Two de-mean modules are added to enhance the interaction between time-series and cross-sectional data 3. The model is trained using IC and weighted MSE loss functions[17] **Model Evaluation**: The model demonstrates improved performance in capturing trends and cross-sectional interactions[17] Quantitative Models Backtesting Results - **DecompGRU TOP200 Portfolio**: - Cumulative absolute return: 38.64% - Excess return relative to WIND All A equal-weight index: 13.8% - Maximum drawdown: 10.08% - Weekly win rate: 64.29% - Monthly win rate: 100% - September absolute return: 4.19% - September excess return: 3.4%[11] - **ETF Rotation Portfolio**: - Cumulative absolute return: 21.54% - Excess return relative to WIND ETF index: -0.57% - Maximum drawdown: 7.82% - Weekly win rate: 65.52% - Monthly win rate: 66.67% - September absolute return: -1.68% - September excess return: -6.65%[13][14] Quantitative Factors and Construction - **Factor Name**: Sentiment Heat Factor **Factor Construction Idea**: The factor aggregates user behavior data (e.g., browsing, self-selection, and clicks) to measure sentiment heat at the stock, index, industry, and concept levels[18] **Factor Construction Process**: 1. Individual stock heat is calculated as the sum of browsing, self-selection, and click counts 2. Normalize the heat value by dividing it by the total market heat on the same day and multiplying by 10,000 3. Aggregate normalized heat values to broader levels such as indices, industries, and concepts[18] **Factor Evaluation**: The sentiment heat factor serves as a proxy for market sentiment and helps identify mispricing due to attention constraints[18] Quantitative Factors Backtesting Results - **Broad Index Sentiment Heat Rotation Strategy**: - Annualized return since 2017: 8.74% - Maximum drawdown: 23.5% - 2025 portfolio return: 32% - Benchmark broad index equal-weight portfolio return: 30%[27] - **Concept Sentiment Heat TOP/BOTTOM Portfolios**: - BOTTOM portfolio annualized return: 15.71% - Maximum drawdown: 28.89% - 2025 BOTTOM portfolio return: 40%[41][45]