Performance analysis of clustering algorithms in medical datasets
P. Premalatha, S. Subasree · 2017 Second International Conference on Electrical, Computer and Communication Technologies (ICECCT) · 2017
Generally, the medical datasets are heterogeneous and large dimensional that contains a million of patient records. Extracting information from such datasets is a tedious process, which can be made easier by some of the clustering algorithms available in data mining. In this paper, three clustering algorithms such as Medical Storage Platform for data Mining (MSPM), Homogeneity Similarity based Hierarchical (HSH) clustering and K-Harmonic Means-Overlapped K-Means (KHM-OKM) clustering is described and their performance is evaluated. The HSH clustering is an enhancement of hierarchical algorithm that considers the homogeneity and relative population of the clusters to measure the Clustering Performance Index (CPI). The MSPM framework is a modification of Apriori algorithm implemented using MapReduce function. This framework enhances the performance of clustering by the parallel processing of Map and Reduce functions. The KHM-OKM clustering is a hybrid algorithm that combines the K-Harmonic Means and Overlapped K-Means clustering. The results of these algorithms are experimentally evaluated regardingCPI, confidence and FBCubed measure.