Modified Particle Swarm Optimization Based Adaptive Fuzzy K-Modes Clustering for Heterogeneous Medical Databases

Abstract

The main purpose of data mining is to extract hidden predictive knowledge of useful information and patterns of data from large databases for utilizing it in decision support. Medical field has large amount of various heterogeneous databases, in which the extraction of hidden useful knowledge for the classification of data is difficult one. In order to cluster and classify the whole databases of medical field, a clustering algorithm MPSO-AFKM (Modified Particle Swarm Optimization based Adaptive Fuzzy K-Modes) is introduced. The proposed method works with the two phases clustering and classification for the effective classification of medical database. The foremost step is the clustering, which utilize the MPSO-AFKM algorithm for obtaining clustered data. In MPSO-AFKM, the categorical data is clustered with Adaptive Fuzzy K-Modes (AFKM) algorithm and the cluster centroids in AFKM is optimized using Modified Particle Swarm Optimization (MPSO) algorithm for getting accurate clustering results. The clustered results of data are classified with the aid of Fuzzy Logic system, by which our required information is achieved. Our proposed work is implemented in Matlab platform on Postoperative Patient dataset. And the performance is also evaluated with the evaluation metrics precision, sensitivity, specificity and accuracy, which shows that our proposed work performance is better one for the effective medical data clustering. Moreover, the comparison is also made to prove the good performance of our proposed work over other existing works.