2024年AI策略回顾：文本量价齐飞，行业轮动出色

Industry Investment Rating - The report highlights strong performance in AI-driven strategies, particularly in text-based stock selection and industry rotation models, indicating a positive outlook for AI-driven investment strategies [1][3][7] Core Views - The text-based stock selection model (FADT_BERT) achieved an absolute return of 23.41% in 2024, with a 17.95% excess return over the CSI 500 Index [1] - The full-frequency fusion factor model outperformed the all-A-share equal-weighted benchmark by 20.29% in 2024, with a 5-day RankIC mean of 0.121 [2] - The AI industry rotation model achieved a 37.49% excess return over the equal-weighted benchmark in 2024, with a historical annualized return of 25.40% since 2017 [3][12] - The AI concept index rotation model achieved a 20.91% annualized return since 2018, with a 12.37% annualized excess return over the equal-weighted benchmark [19] Text-Based Stock Selection Model (FADT_BERT) - The FADT_BERT model has shown consistent performance since 2009, with an annualized return of 40.20% and an annualized excess return of 32.74% over the CSI 500 Index [1][34] - The model's Sharpe ratio stands at 1.40, indicating strong risk-adjusted returns [1] Full-Frequency Fusion Factor Model - The full-frequency fusion factor model has demonstrated robust performance, with a 30.92% annualized excess return since 2017 and a 5-day RankIC mean of 0.115 [2][28] - The AI-enhanced CSI 1000 portfolio achieved a 6.26% absolute return in 2024, with a 5.07% excess return over the CSI 1000 Index [2][28] AI Industry Rotation Model - The AI industry rotation model selects the top 5 industries weekly based on full-frequency price-volume fusion factors, achieving a 37.49% excess return in 2024 [3][12] - The model's historical annualized excess return since 2017 is 22.02%, with a maximum drawdown of 12.43% [12] AI Concept Index Rotation Model - The AI concept index rotation model selects the top 10 concept indices weekly, achieving a 14.23% return in 2024 with a 5.88% excess return over the equal-weighted benchmark [19] - The model's historical annualized return since 2018 is 20.91%, with a 12.37% annualized excess return [19] AI Theme Index Rotation Model - The AI theme index rotation model focuses on thematic indices, achieving a 6.94% return in 2024, though with a slight underperformance of -1.84% relative to the benchmark [24] - The model's historical annualized return since 2018 is 14.72%, with a 12.50% annualized excess return [24] Key Industries and Indices - The AI industry rotation model is currently favoring industries such as banking, real estate, non-bank financials, petrochemicals, and rare metals [3][15] - The AI concept index rotation model is focusing on indices such as software, Wind micro-cap stocks, office supplies, and fintech [19][20] - The AI theme index rotation model is targeting indices like CSI Software, Information Security, and CSI Data [24][25]