A novel hybrid model combining dimensionality reduction techniques and classification algorithms for credit scoring

Credit scoring is a prominent application of machine learning and data mining. Numerous studies have utilized machine learning algorithms, including feature selection, classification, and ensemble learning, within the context of hybrid credit scoring models. However, most of these studies have not compared the extensive variety of compound and hybrid credit scoring methods collectively. This study proposes a novel hybrid credit scoring model that integrates three dimensionality reduction techniques — feature selection, feature extraction, and manifold learning — with two classification approaches: single classifiers and ensemble classifiers. By employing the proposed model, a wide array of hybrid credit scoring models will be developed and evaluated using five performance metrics. The results indicate that the proposed model outperforms other hybrid credit scoring models presented in previous research. This model can be regarded as a comprehensive framework for leveraging machine learning in the development of hybrid credit scoring models.