Estimation of inter-centroid distance quality in data clustering problem using hybridized K-means algorithm

This paper presents, simulate, access and applies the proposed for data classification of medical dataset with aims to classify patients based on medical history. This modified data classification algorithm was formulated using k-Means algorithm. The simulation has been performed by using Real and artificial datasets on MATLAB 7.7.0 and showed that increasing the accuracy of data classification of Diabetes dataset by using proposed hybridization of distance function. The results were compared by using traditional distance function versus proposed distance function for traditional k-Means algorithm. The results shows an improved k-means classification algorithm by applying proposed method for finding minimum distance between centroid. When compared centroid distance by applying traditional Euclidean, Canberra in k-Means algorithm, the proposed modified algorithm proves minimum distance. Simulations shows Canberra distance function perform better as compared to Euclidean and proposed model performs best as compared to Canberra distance function and shows up to 9.29 % improvement over Canberra method for inter centroid distance calculation. We have also run the algorithm and obtain the result showing hybrid distane proves itself with calculating the minimum distance for inter-centroid.