arXiv q-fin· Xintong Liu, Chenkai Zhang, Yanzhao Ma, Peng Yang·· 5 hr agoAI Score45
Evolutionary Factor Mining for Large Language Models Based on Lineage-Aware Search Control
Lineage-Aware Search Control for LLM-Based Evolutionary Factor Mining
AI summary
This paper proposes the LineageEvo method, which optimizes the LLM-driven evolutionary factor mining process through two mechanisms: lineage-conditioned LLM mutation and lineage hierarchical search allocation. The method was evaluated across factor predictability, portfolio performance, search efficiency, and diversity dimensions on the CSI 500 and S&P 500 markets, achieving optimal predictive performance in both markets.
Source:arXiv q-fin · arxiv.org
Academic papers#Papers / Research#Factor Investing#Large Language Model Applications#Machine Learning#A-shares#US Stocks