Portfolio Construction Based on Stock Characteristics Using Function Generation

This paper extends the function-generated portfolio framework by incorporating stock-specific information based on stochastic portfolio theory. It allows portfolio weights to depend on stock characteristics and market weights, demonstrating improved relative arbitrage efficiency and portfolio performance. An empirical analysis using SSE 50 Index data over 10 years shows that, even with transaction costs, portfolios constructed with the ROA indicator outperform market portfolios and other function-generated portfolios, offering superior arbitrage opportunities.