Research on Ensemble Learning Quantitative Investment Model Based on Supply Chain Network Characteristics

In the context of increasing complexity and information interweaving in financial markets, traditional quantitative investment models have encountered problems such as reduced factor effectiveness and inaccurate risk measurement because they do not take into account the supply chain network relationships between enterprises.This paper discusses this issue from the perspective of supply chain networks and creates a quantitative investment model that combines graph structure features with ensemble learning methods.Relying on the upstream and downstream relationships of listed companies, we build an enterprise-level supply chain network, collect multiple dimensional features such as centrality, structural holes, node embedding, etc., detect the potential value transfer and risk diffusion routes between enterprises, and use integrated learning algorithms such as XGBoost and LightGBM to create a multi-model fusion framework to achieve accurate estimation of stock excess returns.Relying on the empirical data of A-shares to carry out strategy backtesting, the effectiveness of the model in terms of return improvement and risk control is confirmed.What significantly improves the explanatory power and robustness of the model is the embedding of the supply chain network, which opens up new horizons for creating a more systematic quantitative investment strategy.This article has built a methodological framework for multi-source financial data integration and network financial modeling, and also provides theoretical support for risk prevention and control and factor innovation in the actual investment process. Introduction Research MotivationIn recent years, as the financial market has been moving towards a highly information-based and