HTSC(SH:601688) HTSC·2025-07-27 10:26Quantitative Models and Construction Methods - Model Name: AI Industry Rotation Model Construction Idea: The model uses full-spectrum price-volume fusion factors to score 32 primary industries and constructs a weekly rebalancing strategy by selecting the top 5 industries for equal-weight allocation [2][23][16] Construction Process: 1. Industry Pool: Includes 32 primary industries, excluding comprehensive and comprehensive finance sectors. Certain industries are split into subcategories (e.g., food and beverage split into food, beverage, and liquor) [23] 2. Factor: Full-spectrum price-volume fusion factor, derived from industry constituent stocks' factor scores [23][16] 3. Strategy Rules: - Select the top 5 industries with the highest scores on the last trading day of each week - Equal-weight allocation - Buy at the closing price of the first trading day of the following week - Weekly rebalancing, no transaction costs considered [23] Evaluation: The model leverages AI's feature extraction capabilities to fully explore patterns in multi-frequency price-volume data, complementing top-down strategies [16] - Model Name: AI Thematic Index Rotation Model Construction Idea: The model uses full-spectrum price-volume fusion factors to score 133 thematic indices and constructs a weekly rebalancing strategy by selecting the top 10 indices for equal-weight allocation [3][9][15] Construction Process: 1. Index Pool: Based on Wind's ETF fund classification, selects indices tracked by thematic ETFs, forming a pool of 133 thematic indices [9] 2. Factor: Full-spectrum price-volume fusion factor, derived from thematic index constituent stocks' factor scores [9][15] 3. Strategy Rules: - Select the top 10 indices with the highest scores on the last trading day of each week - Equal-weight allocation - Buy at the opening price of the first trading day of the following week - Weekly rebalancing, transaction costs set at 0.04% for both sides [9][15] Evaluation: The model effectively identifies high-performing thematic indices using AI-driven factor scoring [15] - Model Name: AI Concept Index Rotation Model Construction Idea: The model uses full-spectrum price-volume fusion factors to score 72 concept indices and constructs a weekly rebalancing strategy by selecting the top 10 indices for equal-weight allocation [11][15][32] Construction Process: 1. Index Pool: Selects 72 popular concept indices from Wind [15] 2. Factor: Full-spectrum price-volume fusion factor, derived from concept index constituent stocks' factor scores [15][32] 3. Strategy Rules: - Select the top 10 indices with the highest scores on the last trading day of each week - Equal-weight allocation - Buy at the opening price of the first trading day of the following week - Weekly rebalancing, transaction costs set at 0.04% for both sides [15][32] Evaluation: The model efficiently captures trends in concept indices using AI-driven factor scoring [32] - Model Name: AI CSI 1000 Enhanced Portfolio Construction Idea: The portfolio is constructed using full-spectrum fusion factors to enhance the CSI 1000 index [1][27][29] Construction Process: 1. Factor: Full-spectrum fusion factor, combining high-frequency deep learning factors and low-frequency multi-task factors [26][29] 2. Portfolio Construction Rules: - Stock weight deviation limit: 0.8 - Barra exposure limit: 0.3 - Weekly turnover rate capped at 30% - Weekly rebalancing, transaction costs set at 0.4% for both sides [29] Evaluation: The portfolio demonstrates strong tracking and enhancement capabilities relative to the CSI 1000 index [27] - Model Name: Text FADT_BERT Stock Selection Portfolio Construction Idea: The portfolio is based on the forecast_adjust_txt_bert factor, which upgrades text factors under earnings forecast adjustment scenarios [32][33][36] Construction Process: 1. Factor: Forecast_adjust_txt_bert factor, derived from text-based analysis of earnings forecast adjustments [32][33] 2. Portfolio Construction Rules: - Enhances the long-side base stock pool - Constructs a top 25 active quantitative stock selection portfolio [32][33] Evaluation: The portfolio effectively integrates text-based AI factors for stock selection [36] Model Backtesting Results - AI Industry Rotation Model: - Annualized return: 25.69% - Annualized excess return: 20.23% - Maximum drawdown of excess return: 12.43% - Excess Sharpe ratio: 1.96 - YTD return: 14.11% - YTD excess return: 0.14% [22] - AI Thematic Index Rotation Model: - Annualized return: 16.65% - Annualized excess return: 12.19% - Maximum drawdown of excess return: 16.55% - Excess Sharpe ratio: 0.96 - YTD return: 16.97% - YTD excess return: 6.87% [8] - AI Concept Index Rotation Model: - Annualized return: 23.67% - Annualized excess return: 12.20% - Maximum drawdown of excess return: 17.96% - Excess Sharpe ratio: 1.03 - YTD return: 23.94% - YTD excess return: 7.06% [13] - AI CSI 1000 Enhanced Portfolio: - Annualized return: 18.95% - Annualized excess return: 22.36% - Annualized tracking error: 6.04% - Maximum drawdown of excess return: 7.55% - Information ratio: 3.70 - Calmar ratio: 2.96 [30] - Text FADT_BERT Stock Selection Portfolio: - Annualized return: 39.73% - Annualized excess return: 31.34% - Maximum drawdown: 48.69% - Sharpe ratio: 1.38 - Calmar ratio: 0.82 [36]