Quantitative Models and Construction Methods 1. Model Name: GRU(10,2)+NN(10) - Model Construction Idea: This model uses a combination of GRU (Gated Recurrent Unit) and Neural Networks to capture high-frequency trading patterns. - Model Construction Process: The model is constructed by first applying a GRU layer with 10 units and 2 layers, followed by a Neural Network layer with 10 units. This combination helps in capturing temporal dependencies and complex patterns in high-frequency data. - Model Evaluation: The model shows stable performance with notable returns. - [3][4] 2. Model Name: GRU(50,2)+NN(10) - Model Construction Idea: Similar to the previous model but with a larger GRU layer to capture more complex patterns. - Model Construction Process: The model uses a GRU layer with 50 units and 2 layers, followed by a Neural Network layer with 10 units. - Model Evaluation: This model also demonstrates strong performance with higher returns compared to the smaller GRU model. - [3][4] 3. Model Name: Multi-Granularity Model (5-day label) - Model Construction Idea: This model uses multi-granularity labels to capture different trading patterns over a 5-day period. - Model Construction Process: The model is trained using a bi-directional AGRU (Attention-based GRU) to capture both past and future dependencies in the data. - Model Evaluation: The model shows significant returns and is particularly effective in capturing short-term trading patterns. - [3][4] 4. Model Name: Multi-Granularity Model (10-day label) - Model Construction Idea: Similar to the 5-day label model but with a 10-day period to capture longer-term patterns. - Model Construction Process: The model is trained using a bi-directional AGRU. - Model Evaluation: This model also shows strong performance, capturing longer-term trading patterns effectively. - [3][4] Model Backtesting Results 1. GRU(10,2)+NN(10) - Multi-Period Return: 0.84% (last week), 0.84% (April 2025), 9.81% (YTD 2025) - [3] 2. GRU(50,2)+NN(10) - Multi-Period Return: 0.87% (last week), 0.87% (April 2025), 11.21% (YTD 2025) - [3] 3. Multi-Granularity Model (5-day label) - Multi-Period Return: 1.43% (last week), 1.43% (April 2025), 22.01% (YTD 2025) - [3] 4. Multi-Granularity Model (10-day label) - Multi-Period Return: 1.28% (last week), 1.28% (April 2025), 19.89% (YTD 2025) - [3] Quantitative Factors and Construction Methods 1. Factor Name: Intraday Skewness Factor - Factor Construction Idea: This factor captures the skewness in intraday returns to identify potential trading opportunities. - Factor Construction Process: The factor is calculated based on the distribution of intraday returns, focusing on the skewness measure. - Factor Evaluation: The factor shows mixed performance with some periods of negative returns. - [3][4] 2. Factor Name: Downside Volatility Proportion Factor - Factor Construction Idea: This factor measures the proportion of downside volatility in intraday trading. - Factor Construction Process: The factor is calculated by decomposing realized volatility into upside and downside components and focusing on the downside proportion. - Factor Evaluation: The factor shows relatively stable performance with positive returns in most periods. - [3][4] 3. Factor Name: Post-Open Buy Intention Proportion Factor - Factor Construction Idea: This factor measures the proportion of buy intentions after the market opens. - Factor Construction Process: The factor is calculated based on the proportion of buy orders in the initial trading period after the market opens. - Factor Evaluation: The factor shows consistent positive returns, indicating strong buy signals. - [3][4] 4. Factor Name: Post-Open Buy Intensity Factor - Factor Construction Idea: This factor measures the intensity of buy intentions after the market opens. - Factor Construction Process: The factor is calculated based on the intensity and volume of buy orders in the initial trading period after the market opens. - Factor Evaluation: The factor shows stable performance with positive returns. - [3][4] Factor Backtesting Results 1. Intraday Skewness Factor - Multi-Period Return: -0.12% (last week), -0.12% (April 2025), 11.2% (YTD 2025) - [3] 2. Downside Volatility Proportion Factor - Multi-Period Return: -0.36% (last week), -0.36% (April 2025), 8.82% (YTD 2025) - [3] 3. Post-Open Buy Intention Proportion Factor - Multi-Period Return: 0.41% (last week), 0.41% (April 2025), 6.53% (YTD 2025) - [3] 4. Post-Open Buy Intensity Factor - Multi-Period Return: 0.76% (last week), 0.76% (April 2025), 7% (YTD 2025) - [3]
高频选股因子周报(20250331- 20250403):上周大单因子表现优异,中证 1000 AI 增强严约束组合尤为强势-20250410
国泰海通证券·2025-04-10 07:22