Credit Scoring Prediction for Small and Medium-Sized Enterprises Based on Alternative Data and Gradient Boosting Algorithms

In today's complex economic landscape, small and medium-sized enterprises (SMEs) are crucial drivers of growth, yet traditional credit scoring models often fail to capture their true creditworthiness because they are limited by narrow data sources and poor data adaptability. With the rise of big data and fintech, alternative data opens a richer avenue for SME credit assessment. This study leverages real-world, publicly available data, which includes operational behavior, supply chain interactions, and online transactions, to help build a more inclusive and forward-looking credit scoring framework for SMEs. The authors enhance the model's nonlinear fitting and feature representation capabilities by employing gradient boosting algorithms to significantly improve credit risk prediction accuracy. They compare the performance of various machine learning models and discuss trade-offs between predictive power, generalizability, and interpretability. The results offer financial institutions a dynamic, multidimensional risk assessment tool able to provide actionable insights for policy and practice.