Application of explainable machine learning based on Catboost in credit scoring

Abstract

Credit scoring is the core part of an institution's lending. As artificial intelligence is used in various fields, credit rating is also under the same topic of accepting technological changes. Combining credit evaluation and machine learning can incorporate relatively comprehensive features into the credit evaluation process. Through the excellent performance of Catboost, while ensuring accuracy, it demonstrates the explainability of the model as much as possible, avoiding the traditional trust problem of the black-box model. Explainability is proposed to the machine learning model, which reduces the difficulty of processing large amounts of data and the threshold for non-professionals to understand the model. In this article, the dataset is the personal loan data of LendingClub obtained through python. By analyzing the data through Catboost, we can derive excellent results in applying the explainability of machine learning in personal credit evaluation.