Uncovering Chronic Conditions using Featured-based Clustering Approach

Abstract Chronic conditions like diabetes and hypertension are prevalent worldwide, impacting a large portion of the population. It’s imperative to develop more efficient healthcare models to manage the burden of chronic diseases. Data-driven cluster analysis and its proper validation is an important problem in healthcare and other application domains. This study aimes to discover homogenous groups of people with chronic conditions based on a feature-based and statistical cluster validation approach. The featured-based approach uses information from raw data features and cross-validation to validate the optimal clustering results. Moreover, the distinctness of clusters was assessed using statistical measures in terms of variables where statistically significant results were obtained between the clusters. The experimental results have shown that the proposed approach is effective and accurate in determining the optimal number of clustering groups. The discovered clusters can serve as a robust basis for interpreting potential correlations among patient characteristics within the context of chronic conditions and other factors.