从微观出发的风格轮动月度跟踪-20251201

Quantitative Models and Construction Methods - Model Name: Style Rotation Model Model Construction Idea: The model is built from basic style factors such as valuation, market capitalization, volatility, and momentum. It incorporates a style timing and scoring system, leveraging micro-level features and machine learning techniques to optimize style selection and rotation[4][9]. Model Construction Process: 1. Start with 80 fundamental micro factors as raw features, categorized based on the proprietary multi-factor system of Dongwu Securities[9]. 2. Construct 640 micro-level features from these factors[4][9]. 3. Replace the absolute proportion division of style factors with commonly used indices as style stock pools, creating new style returns as labels[4][9]. 4. Use a rolling training process with a Random Forest model to avoid overfitting, select optimal features, and generate style recommendations[4][9]. 5. Combine style timing results and scoring outcomes to build a monthly frequency style rotation framework[4][9]. Model Evaluation: The model effectively integrates micro-level features and machine learning to enhance style rotation performance, mitigating overfitting risks[4][9]. Model Backtesting Results - Style Rotation Model: - Annualized Return: 16.52% - Annualized Volatility: 20.46% - IR: 0.81 - Monthly Win Rate: 57.01% - Maximum Drawdown: 25.68% - Excess Annualized Return (Hedged against Benchmark): 11.04% - Excess Annualized Volatility (Hedged against Benchmark): 11.08% - Excess IR (Hedged against Benchmark): 1.00 - Excess Monthly Win Rate (Hedged against Benchmark): 55.14% - Maximum Drawdown (Hedged against Benchmark): 9.00%[4][10][11]