金融工程 证 券 研 究 报 告 【金工周报】（20260224-20260227） 短期模型以中性为主，后市或维持中性震荡 本周回顾 本周市场普遍上涨，上证指数单周上涨 1.98%，创业板指单周上涨 1.05%。 A 股模型： 短期：成交量模型中性。特征龙虎榜机构模型中性。特征成交量模型看空。智 能算法沪深 300 模型中性，智能算法中证 500 模型中性。 中期：涨跌停模型中性。上下行收益差模型绝大部分宽基指数看多。月历效应 模型中性。 长期：长期动量模型中性。 综合：A 股综合兵器 V3 模型看空。A 股综合国证 2000 模型看空。 港股模型： 中期：成交额倒波幅模型看空。上下行收益差模型中性，上下行收益差相似模 型看多。 本周行业指数普遍上涨，涨幅前五的行业为：钢铁、有色金属、基础化工、煤 炭、电力及公用事业，跌幅前五的行业为：传媒、消费者服务、食品饮料、非 银行金融、银行。从资金流向角度来说，有色金属、石油石化、银行、农林牧 渔主力资金净流入居前，传媒、电力设备及新能源、计算机、机械、基础化工 主力资金净流出居前。 本周股票型基金总仓位为 95.75%，相较于上周增加了 24 个 bps，混合型基 ...

[Table_Summary] 报告摘要： | [分析师： Table_Author]刘晨明 | | --- | | SAC 执证号：S0260524020001 | | SFC CE No. BVH021 | | 010-59136616 | | liuchenming@gf.com.cn | | 分析师： 郑恺 | | SAC 执证号：S0260515090004 | | SFC CE No. BUU989 | | 021-38003559 | | zhengkai@gf.com.cn | | 分析师： 李如娟 | | SAC 执证号：S0260524030002 | | 020-66336563 | | lirujuan@gf.com.cn | | 请注意，李如娟并非香港证券及期货事务监察委员会的注 | | 册持牌人，不可在香港从事受监管活动。 | [Table_ 相关研究： DocReport] | 假期非美市场延续牛市氛 | 2026-02-23 | | --- | --- | | 围:——春节大事 5 分钟全知 | | | 道 | | | 春节前后，港股如何反 | 2026-02-08 | | 应 ...

- The short-term trading volume model is neutral[2][11] - The characteristic institutional model based on the Dragon and Tiger list is neutral[2][11] - The characteristic trading volume model is bearish[2][11] - The intelligent algorithm model for the CSI 300 is bullish[2][11] - The intelligent algorithm model for the CSI 500 is bearish[2][11] - The mid-term limit-up and limit-down model is neutral[2][12] - The mid-term up-down return difference model is bullish for some broad-based indices[2][12] - The mid-term calendar effect model is bullish[2][12] - The long-term momentum model is neutral[2][12] - The comprehensive A-share V3 model is neutral[2][13] - The comprehensive A-share Guozheng 2000 model is neutral[2][13] - The mid-term trading volume to volatility model for Hong Kong stocks is bearish[2][13] - The Hang Seng Index up-down return difference model is neutral[2][13] - The Hang Seng Index up-down return similarity model is bullish[2][13]

- The short-term trading volume model indicates a bullish outlook for some broad-based indices [1][10] - The characteristic institutional model from the Dragon and Tiger list is neutral [1][10] - The characteristic trading volume model is neutral [1][10] - The intelligent algorithm model for the CSI 300 index is bullish, while the intelligent algorithm model for the CSI 500 index is bearish [1][10] - The mid-term limit-up and limit-down model is neutral [1][11] - The up-and-down return difference model is bullish for all broad-based indices [1][11] - The calendar effect model is neutral [1][11] - The long-term momentum model is neutral [1][12] - The comprehensive A-share V3 model is bullish [1][13] - The comprehensive A-share Guozheng 2000 model is neutral [1][13] - The mid-term trading volume to volatility model for Hong Kong stocks is bullish [1][14] - The up-and-down return difference model for the Hang Seng Index is neutral, while the similar up-and-down return difference model is bullish [1][14]

Core Viewpoint - The article emphasizes the importance of avoiding risky investment behaviors and highlights the principle of missing out on opportunities rather than making wrong investment decisions [2]. Group 1: Investment Risks - Recent declines in commercial aerospace, AI applications, and robotics stocks have exceeded 40% from their peaks, indicating potential financial disasters for investors who used leverage or bought into unfamiliar assets [2]. - Prominent financial influencers have faced penalties for promoting speculative investments, leading to significant losses for investors who followed their advice [2]. - Investors are advised to create a negative checklist to avoid known risks, focusing on what not to do rather than what to do [2]. Group 2: Wealth Accumulation - The pursuit of quick wealth is discouraged, as it often leads to gambling-like behavior rather than sound investment practices [3]. - Wealth accumulation requires time, patience, knowledge, discipline, and hard work, and those promoting easy riches should be approached with caution [3]. - The article references Charlie Munger's view that seeking quick wealth can lead to negative traits like jealousy and arrogance, resulting in poor investment decisions [4]. Group 3: Common Investment Mistakes - Avoid trying to discover the next big company like Microsoft; instead, focus on reliable companies whose stock prices are undervalued [5]. - The belief that "this time is different" in the market is a costly lesson, as historical patterns tend to repeat themselves [6]. - Investors should not let personal preferences for a company's products cloud their judgment regarding its profitability [6]. - Panic selling during market downturns is discouraged, as stocks are often most attractive when no one wants to buy them [6]. - Timing the market is deemed an investment myth, with no strategy consistently predicting the best times to buy or sell [7]. - Valuation should not be overlooked; investments should be based on current value rather than speculation on future buyers [7]. - Cash flow analysis is crucial for assessing a company's financial health, as it provides a clearer picture than earnings figures alone [7].

Quantitative Models and Construction - **Model Name**: Timing Radar Six-Dimensional Framework **Model Construction Idea**: The model integrates multi-dimensional indicators to assess equity market performance, categorizing them into four major dimensions: "Valuation Cost-Effectiveness," "Macro Fundamentals," "Funds & Trends," and "Crowdedness & Reversal," generating a composite timing score within [-1,1] range [1][6][9] **Model Construction Process**: 1. Select 21 indicators across liquidity, economic fundamentals, valuation, funds, technical trends, and crowdedness dimensions [1][6] 2. Aggregate these indicators into four categories: "Valuation Cost-Effectiveness," "Macro Fundamentals," "Funds & Trends," and "Crowdedness & Reversal" [1][6] 3. Calculate a composite timing score within the range [-1,1] based on the aggregated indicators [1][6] **Model Evaluation**: Provides a comprehensive view of market conditions, but its effectiveness depends on stable market environments [1][6] Quantitative Factors and Construction Liquidity Factors - **Factor Name**: Monetary Direction Factor **Factor Construction Idea**: Measures the direction of monetary policy based on changes in policy rates and short-term market rates [12] **Factor Construction Process**: 1. Calculate the average change in policy rates and short-term market rates over the past 90 days 2. If the factor > 0, monetary policy is considered loose; if < 0, monetary policy is considered tight [12] **Factor Evaluation**: Effectively captures monetary policy trends [12] - **Factor Name**: Monetary Strength Factor **Factor Construction Idea**: Represents the deviation of short-term market rates from policy rates using the "interest rate corridor" concept [15] **Factor Construction Process**: 1. Compute deviation = DR007/7-year reverse repo rate - 1 2. Smooth and standardize the deviation using z-score 3. Assign scores: <-1.5 SD = 1 (loose environment), >1.5 SD = -1 (tight environment) [15] **Factor Evaluation**: Captures short-term liquidity deviations effectively [15] - **Factor Name**: Credit Direction Factor **Factor Construction Idea**: Reflects the transmission of credit from banks to the real economy [18] **Factor Construction Process**: 1. Use monthly long-term loan data 2. Calculate past 12-month increments and year-over-year changes 3. Compare with three months prior: upward trend = 1, downward trend = -1 [18] **Factor Evaluation**: Provides insights into credit transmission trends [18] - **Factor Name**: Credit Strength Factor **Factor Construction Idea**: Measures whether credit indicators significantly exceed or fall short of expectations [21] **Factor Construction Process**: 1. Compute deviation = (new RMB loans - median forecast)/forecast SD 2. Assign scores: >1.5 SD = 1 (credit exceeds expectations), <-1.5 SD = -1 (credit falls short) [21] **Factor Evaluation**: Captures unexpected credit changes effectively [21] Economic Factors - **Factor Name**: Growth Direction Factor **Factor Construction Idea**: Based on PMI data to assess economic growth trends [23] **Factor Construction Process**: 1. Use PMI data (manufacturing, non-manufacturing, Caixin manufacturing) 2. Calculate past 12-month averages and year-over-year changes 3. Compare with three months prior: upward trend = 1, downward trend = -1 [23] **Factor Evaluation**: Reflects economic growth trends effectively [23] - **Factor Name**: Growth Strength Factor **Factor Construction Idea**: Measures whether economic growth indicators significantly exceed or fall short of expectations [26] **Factor Construction Process**: 1. Compute deviation = (PMI - median forecast)/forecast SD 2. Assign scores: >1.5 SD = 1 (growth exceeds expectations), <-1.5 SD = -1 (growth falls short) [26] **Factor Evaluation**: Captures unexpected economic growth changes effectively [26] - **Factor Name**: Inflation Direction Factor **Factor Construction Idea**: Assesses inflation trends based on CPI and PPI data [28] **Factor Construction Process**: 1. Compute inflation direction = 0.5 × smoothed CPI YoY + 0.5 × raw PPI YoY 2. Compare with three months prior: downward trend = 1, upward trend = -1 [28] **Factor Evaluation**: Reflects inflation trends effectively [28] - **Factor Name**: Inflation Strength Factor **Factor Construction Idea**: Measures whether inflation indicators significantly exceed or fall short of expectations [30] **Factor Construction Process**: 1. Compute deviation = (CPI/PPI - median forecast)/forecast SD 2. Assign scores: <-1.5 SD = 1 (inflation falls short), >1.5 SD = -1 (inflation exceeds expectations) [30] **Factor Evaluation**: Captures unexpected inflation changes effectively [30] Valuation Factors - **Factor Name**: Shiller ERP **Factor Construction Idea**: Adjusts earnings for inflation to assess market valuation [31] **Factor Construction Process**: 1. Compute Shiller PE = average inflation-adjusted earnings over the past six years 2. Calculate Shiller ERP = 1/Shiller PE - 10-year bond yield 3. Standardize using z-score over the past six years [31] **Factor Evaluation**: Provides a robust valuation metric [31] - **Factor Name**: PB **Factor Construction Idea**: Standardizes PB to assess market valuation [35] **Factor Construction Process**: 1. Compute PB × (-1) 2. Standardize using z-score over the past six years 3. Truncate at ±1.5 SD and normalize to ±1 range [35] **Factor Evaluation**: Effectively captures valuation trends [35] - **Factor Name**: AIAE **Factor Construction Idea**: Reflects market-wide equity allocation and risk appetite [37] **Factor Construction Process**: 1. Compute AIAE = total equity market cap/(total equity market cap + total debt) 2. Multiply by (-1) and standardize using z-score over the past six years [37] **Factor Evaluation**: Captures market risk appetite effectively [37] Funds Factors - **Factor Name**: Margin Trading Increment **Factor Construction Idea**: Measures market leverage trends [40] **Factor Construction Process**: 1. Compute margin balance - short balance 2. Compare 120-day average increment with 240-day average increment: upward trend = 1, downward trend = -1 [40] **Factor Evaluation**: Reflects leverage trends effectively [40] - **Factor Name**: Turnover Trend **Factor Construction Idea**: Measures market activity and liquidity [43] **Factor Construction Process**: 1. Compute log turnover moving average distance = ma120/ma240 - 1 2. Assign scores: max(10/30/60) = 1, min(10/30/60) = -1 [43] **Factor Evaluation**: Captures market activity effectively [43] - **Factor Name**: China Sovereign CDS Spread **Factor Construction Idea**: Reflects foreign investors' sentiment toward China's credit risk [47] **Factor Construction Process**: 1. Compute smoothed CDS spread 20-day difference 2. Assign scores: <0 = 1 (positive sentiment), >0 = -1 (negative sentiment) [47] **Factor Evaluation**: Captures foreign sentiment effectively [47] - **Factor Name**: Overseas Risk Aversion Index **Factor Construction Idea**: Reflects global market risk appetite [49] **Factor Construction Process**: 1. Compute smoothed risk aversion index 20-day difference 2. Assign scores: <0 = 1 (positive sentiment), >0 = -1 (negative sentiment) [49] **Factor Evaluation**: Captures global risk appetite effectively [49] Technical Factors - **Factor Name**: Price Trend **Factor Construction Idea**: Measures market trend direction and strength [52] **Factor Construction Process**: 1. Compute moving average distance = ma120/ma240 - 1 2. Assign scores: >0 = 1 (upward trend), <0 = -1 (downward trend) 3. Combine trend direction and strength scores [52] **Factor Evaluation**: Captures market trends effectively [52] - **Factor Name**: New Highs and Lows **Factor Construction Idea**: Reflects reversal signals based on constituent stocks' highs and lows [54] **Factor Construction Process**: 1. Compute smoothed new lows - new highs

Quantitative Models and Construction Methods 1. Model Name: Volume Model - **Construction Idea**: The model uses trading volume data to predict market trends[1][13] - **Construction Process**: The model analyzes the trading volume of various broad-based indices to generate buy or sell signals[1][13] - **Evaluation**: The model is effective in capturing short-term market movements[1][13] 2. Model Name: Feature Dragon Tiger List Institutional Model - **Construction Idea**: This model uses institutional trading data from the Dragon Tiger List to predict market trends[1][13] - **Construction Process**: The model analyzes the trading activities of institutions listed on the Dragon Tiger List to generate buy or sell signals[1][13] - **Evaluation**: The model is useful for understanding institutional trading behavior and its impact on the market[1][13] 3. Model Name: Feature Volume Model - **Construction Idea**: This model uses specific volume characteristics to predict market trends[1][13] - **Construction Process**: The model analyzes specific volume patterns to generate buy or sell signals[1][13] - **Evaluation**: The model is effective in identifying significant volume changes that precede market movements[1][13] 4. Model Name: Intelligent Algorithm CSI 300 Model - **Construction Idea**: This model uses intelligent algorithms to predict the CSI 300 index trends[1][13] - **Construction Process**: The model employs machine learning algorithms to analyze historical data and generate buy or sell signals for the CSI 300 index[1][13] - **Evaluation**: The model leverages advanced algorithms to improve prediction accuracy[1][13] 5. Model Name: Intelligent Algorithm CSI 500 Model - **Construction Idea**: This model uses intelligent algorithms to predict the CSI 500 index trends[1][13] - **Construction Process**: The model employs machine learning algorithms to analyze historical data and generate buy or sell signals for the CSI 500 index[1][13] - **Evaluation**: The model leverages advanced algorithms to improve prediction accuracy[1][13] 6. Model Name: Limit Up and Down Model - **Construction Idea**: This model uses the occurrence of limit up and down events to predict market trends[1][13] - **Construction Process**: The model analyzes the frequency and context of limit up and down events to generate buy or sell signals[1][13] - **Evaluation**: The model is effective in capturing extreme market movements[1][13] 7. Model Name: Up and Down Return Difference Model - **Construction Idea**: This model uses the difference between upward and downward returns to predict market trends[1][13] - **Construction Process**: The model calculates the difference between upward and downward returns to generate buy or sell signals[1][13] - **Evaluation**: The model provides insights into market momentum and potential reversals[1][13] 8. Model Name: Calendar Effect Model - **Construction Idea**: This model uses calendar-based patterns to predict market trends[1][13] - **Construction Process**: The model analyzes historical data to identify recurring calendar-based patterns and generate buy or sell signals[1][13] - **Evaluation**: The model is useful for identifying seasonal trends in the market[1][13] 9. Model Name: Long-term Momentum Model - **Construction Idea**: This model uses long-term momentum to predict market trends[1][14] - **Construction Process**: The model analyzes long-term price momentum to generate buy or sell signals[1][14] - **Evaluation**: The model is effective in capturing long-term market trends[1][14] 10. Model Name: A-Share Comprehensive Weapon V3 Model - **Construction Idea**: This model combines multiple factors to predict market trends[1][15] - **Construction Process**: The model integrates various indicators and models to generate a comprehensive buy or sell signal[1][15] - **Evaluation**: The model provides a holistic view of the market by combining multiple factors[1][15] 11. Model Name: A-Share Comprehensive Guozheng 2000 Model - **Construction Idea**: This model combines multiple factors to predict the Guozheng 2000 index trends[1][15] - **Construction Process**: The model integrates various indicators and models to generate a comprehensive buy or sell signal for the Guozheng 2000 index[1][15] - **Evaluation**: The model provides a holistic view of the market by combining multiple factors[1][15] 12. Model Name: Turnover Rate Inverse Volatility Model - **Construction Idea**: This model uses the inverse relationship between turnover rate and volatility to predict market trends[1][16] - **Construction Process**: The model analyzes the turnover rate and its inverse relationship with volatility to generate buy or sell signals[1][16] - **Evaluation**: The model is effective in identifying periods of high market uncertainty[1][16] Model Backtesting Results 1. Volume Model - **Indicator Value**: All broad-based indices are bullish[1][13] 2. Feature Dragon Tiger List Institutional Model - **Indicator Value**: Bullish[1][13] 3. Feature Volume Model - **Indicator Value**: Bullish[1][13] 4. Intelligent Algorithm CSI 300 Model - **Indicator Value**: Bullish[1][13] 5. Intelligent Algorithm CSI 500 Model - **Indicator Value**: Bullish[1][13] 6. Limit Up and Down Model - **Indicator Value**: Bullish[1][13] 7. Up and Down Return Difference Model - **Indicator Value**: All broad-based indices are bullish[1][13] 8. Calendar Effect Model - **Indicator Value**: Neutral[1][13] 9. Long-term Momentum Model - **Indicator Value**: Some broad-based indices are bullish[1][14] 10. A-Share Comprehensive Weapon V3 Model - **Indicator Value**: Bullish[1][15] 11. A-Share Comprehensive Guozheng 2000 Model - **Indicator Value**: Bullish[1][15] 12. Turnover Rate Inverse Volatility Model - **Indicator Value**: Bearish[1][16]

Core Viewpoint - The article suggests that despite market-timers' aversion to bonds, it may be an opportune time to consider investing in them due to potential market shifts and changing economic conditions [1] Group 1: Market Sentiment - Market-timers generally dislike bonds, viewing them as less attractive compared to equities [1] - Current economic indicators suggest a potential shift that could make bonds more appealing [1] Group 2: Investment Opportunities - The article highlights that bonds may offer a safer investment alternative in the current market environment [1] - With interest rates potentially stabilizing, the yield on bonds could become more attractive to investors [1]

- Short-term volume models for some broad-based indices turned bullish[1][3][11] - Feature-based institutional model turned bullish[1][3][11] - Feature-based volume model remained neutral[1][3][11] - Intelligent algorithm model for CSI 300 remained neutral, while for CSI 500 turned bullish[1][3][11] - Mid-term limit-up and limit-down model turned bullish[1][3][12] - Up-down return difference model turned bullish for all broad-based indices[1][3][12] - Calendar effect model remained neutral[1][3][12] - Long-term momentum model turned bullish for some broad-based indices[1][3][13] - Comprehensive A-share V3 model turned bullish[1][3][13] - Comprehensive A-share Guozheng 2000 model turned bullish[1][3][13] - Mid-term turnover amplitude model for Hong Kong stocks turned bullish[1][3][14] - Hang Seng Index up-down return difference model remained neutral[1][3][14]

- The report discusses multiple quantitative models for market timing, including short-term, medium-term, and long-term models. These models are constructed based on principles such as price-volume relationships, momentum, and calendar effects. The short-term models include the "Volume Model," "Feature Institutional Model," and "Feature Volume Model," while medium-term models include the "Limit-Up/Down Model" and "Up/Down Return Difference Model." The long-term model is the "Long-Term Momentum Model"[8][11][12][13] - The construction process of these models involves combining signals from different time horizons and strategies. For example, the "Volume Model" evaluates market activity through trading volume, while the "Momentum Model" focuses on price trends. The "Limit-Up/Down Model" identifies market sentiment by analyzing the frequency of limit-up and limit-down events. The "Up/Down Return Difference Model" measures the difference between upward and downward returns to gauge market direction[8][11][12] - The evaluation of these models suggests that combining signals from different models enhances robustness. For instance, some models are defensive, while others are aggressive, allowing for a balanced approach. The report emphasizes that simplicity in model design often leads to better generalization and performance[8][11][12] - Backtesting results for these models indicate varying levels of effectiveness. For example, the "Long-Term Momentum Model" is currently bullish, while the "Up/Down Return Difference Model" shows a positive outlook across all broad-based indices. The "Feature Institutional Model" is bullish, whereas the "Feature Volume Model" is bearish. The "Volume Model" remains neutral across all indices[11][12][13]