Predictive analysis using hybrid clustering in diabetes diagnosis

Data mining has become crucial in the health care domain for the purpose of predictive analysis. With the discovery of new models, it has become easier to analyze the vast amount of data available in the medical industry. In this research work, K*-Means has been used for removal of the inconsistency found in the data and for optimal feature selection genetic algorithm is used with SVM for the purpose of classification. K*-Means is an optimized hierarchical clustering method which aims at reduction of computational cost. The application of the proposed hybrid clustering model applied on Pima Indians Diabetes dataset shows increase in accuracy by 1.351% and in both sensitivity and positive predicted value by 2.0411%. The proposed model attains better results in comparison to the already existing models in the literature.