Traditional credit risk evaluation methods for student loans rely heavily on expert empirical knowledge, leading to limitations in handling uneven sample distribution and indicator redundancy, which makes it difficult to adapt to the increasingly complex credit changes of student loans. To address these problems, this paper proposes a credit risk evaluation and parameter reduction method for college student loans based on Rough Set (RS) theory. First, a preliminary credit risk evaluation index system consisting of 26 indicators was constructed, and each indicator was coded and quantified. On this basis, a five-tuple decision information system for student loan credit risk assessment was established based on classical rough set theory, and the concepts of attribute dependence and attribute importance were defined, and a dimensionality reduction algorithm is designed. The experimental study was conducted using 6500 groups of student loan data from a university in Shaanxi Province from 2010 to 2025. The KNN algorithm, which exhibited the best classification performance in the comparative test of multiple algorithms, was selected as the classification model to verify the performance of the reduced parameter sets. The experimental results show that the D4 parameter set with 18 reduced indicators achieves the optimal evaluation effect: the classification accuracy reaches 97.14%, the mean square error is only 0.0286, and the symmetric mean absolute percentage error is merely 1.14%, which are significantly better than the original data set and other reduced sets.

college student loan, credit risk assessment, rough set, parameter reduction, KNN algorithm