Abstract: This paper introduces an innovative hybrid unsupervised classification method, termed the Two-Stage Hybrid System with Spectral Clustering and Semi-Supervised Support Vector Machine (TSC-SVM). This method adeptly tackles the unsupervised imbalance problem prevalent in credit risk assessment by focusing on global optimal solutions. Additionally, we present a multi-view combined unsupervised approach that effectively mines data for deeper insights and bolsters the robustness of label predictions by harmonizing discrepancies across three different perspectives.