Intelligent Segmentation of Cooperative Members Using a Hybrid Clustering Approach

Abstract

Member management in savings and loan cooperatives faces challenges due to heterogeneous member characteristics and increasing credit risk. Member segmentation is essential for data-driven decision-making; however, the mixed nature of cooperative data, which consist of both numerical and categorical attributes, limits the effectiveness of single clustering methods. This study proposes a hybrid clustering framework that sequentially integrates K-Means and K-Modes to generate more comprehensive and interpretable member segments. K-Means is first applied to identify patterns in numerical attributes, whereas the resulting cluster labels are incorporated into K-Modes to enhance the clustering of categorical attributes. The optimal number of clusters was determined using the Elbow Method and Silhouette Analysis. A case study was conducted using 3,216 member records from a savings and loan cooperative, containing demographic, financial, and transactional characteristics. The experimental results indicate that K-Means produced three optimal clusters with a silhouette score of 0.237, while the hybrid framework generated four final member segments with clearer operational interpretations: loyal high-value members, low-risk growth members, seasonal agribusiness members, and economically vulnerable members. The findings demonstrate that the proposed hybrid clustering approach provides more comprehensive and actionable segmentation than single-method clustering, thereby supporting cooperative credit risk management, service development, and data-driven decision-making.