金融衍生品周度报告：期债长周期回升

Quantitative Models and Construction 1. Model Name: Short-Cycle Model - Model Construction Idea: Focuses on high-frequency financial data, including market style, external factors, and liquidity[16] - Model Construction Process: - Selects effective and relatively independent factors from high-dimensional data - Incorporates subjective analysis frameworks to build models with out-of-sample generalization capabilities - Signals are derived from the weighted combination of three independent models, with signal strength ranging from 0 to 1[16][17] - Formula for signal combination: $\text{Comprehensive Signal Strength} = w_1 \cdot \text{Model 1 Signal} + w_2 \cdot \text{Model 2 Signal} + w_3 \cdot \text{Model 3 Signal}$ - Positions are determined based on signal thresholds: - Long positions for signals ≥ 0.6 - Short positions for signals ≤ 0.4[17] - Model Evaluation: Focuses on high-frequency data and provides actionable signals for short-term trading[16] 2. Model Name: Long-Cycle Model - Model Construction Idea: Focuses on low-frequency macroeconomic data and market expectations[16] - Model Construction Process: - Similar to the short-cycle model, selects independent factors from high-dimensional data - Incorporates macroeconomic indicators to capture long-term trends - Signals are combined with weights similar to the short-cycle model[16][17] - Model Evaluation: Provides insights into long-term market trends and complements the short-cycle model[16] 3. Model Name: Cross-Asset Arbitrage Strategy (N-S Model) - Model Construction Idea: Combines a fundamental three-factor model with a trend regression model to generate trading signals[22] - Model Construction Process: - Fundamental model based on Nelson-Siegel instantaneous forward rate function: $\mathbf{R}(t) = \beta_{0} + \beta_{1}\frac{1-e^{-t/\tau}}{t/\tau} + \beta_{2}\left(\frac{1-e^{-t/\tau}}{t/\tau} - e^{-t/\tau}\right)$ - $\beta_0$: Level factor, representing market expectations of forward rates - $\beta_1$: Slope factor, representing bond risk premiums - $\beta_2$: Curvature factor, representing convexity deviations - Principal Component Analysis (PCA) and logistic regression are used to classify signals into three categories: - '1': Large spread likely to decrease - '0': Uncertain or oscillating spread - '-1': Large spread likely to increase - Trend regression model filters signals, and trades are executed when signals resonate - Duration-neutral adjustments are made for 10Y-5Y spreads with a 1:1.8 ratio[22] - Model Evaluation: Combines fundamental and trend-based approaches, enhancing signal reliability[22] --- Backtesting Results of Models 1. Short-Cycle Model - Comprehensive Signal Strength: - IF: 0.53 - IH: 0.52 - IC: 0.53 - IM: 0.55 - T: 0.53 - TF: 0.54[17] 2. Long-Cycle Model - Comprehensive Signal Strength: - IF: 0.52 - IH: 0.51 - IC: 0.53 - IM: 0.54 - T: 0.50 - TF: 0.48[17] 3. Cross-Asset Arbitrage Strategy (N-S Model) - Trading Signals: - May 27: -1 (N-S Model), 0 (Trend Regression) - May 28: 0 (N-S Model), 0 (Trend Regression) - May 29: 0 (N-S Model), 0 (Trend Regression) - May 30: 0 (N-S Model), 0 (Trend Regression) - May 31: -1 (N-S Model), 0 (Trend Regression)[25] --- Quantitative Factors and Construction 1. Factor Name: Inflation Indicators - Factor Construction Idea: Measures inflationary pressures using commodity prices and indices[3] - Factor Construction Process: - Includes metrics such as vegetable basket price index, refined copper prices, and natural gas import prices - Historical percentiles and correlations with stock and bond indices are calculated[3] - Factor Evaluation: Captures inflation trends and their impact on financial markets[3] 2. Factor Name: Liquidity Indicators - Factor Construction Idea: Tracks short-term liquidity conditions using interbank rates and the USD index[4] - Factor Construction Process: - Includes DR007, DR001, SHIBOR, and USD index - Historical percentiles and correlations with stock and bond indices are calculated[4] - Factor Evaluation: Reflects the availability of liquidity in the financial system[4] 3. Factor Name: Market Sentiment Indicators - Factor Construction Idea: Measures investor sentiment using financing balances and trading volumes[6][7] - Factor Construction Process: - Stock market sentiment: Financing balances, margin trading balances, and net purchases via Stock Connect - Bond market sentiment: 10Y government bond yields, credit spreads, and trading volumes[6][7] - Factor Evaluation: Provides insights into market risk appetite and sentiment shifts[6][7] --- Backtesting Results of Factors 1. Inflation Indicators - Historical Percentiles: - Vegetable Basket Index: 0.05 - Refined Copper Prices: 0.98 - Natural Gas Import Prices: 0.40[3] 2. Liquidity Indicators - Historical Percentiles: - DR007: 0.45 - DR001: 0.68 - USD Index: 0.70[4] 3. Market Sentiment Indicators - Historical Percentiles: - Stock Market Sentiment: Financing Balance: 0.27, Margin Trading Balance: 0.00 - Bond Market Sentiment: 10Y Government Bond Yield: 0.04, Credit Spread: 0.29[6][7]