An Improved K-Means Clustering Algorithm For Pattern Discovery In Data Mining

Abstract

In health monitoring systems enormous amount of information has been extricated by media sensors, with the assist of medical diagnosis which produces text, audio, video images (media contents). The traditional method provides immense process lots of complexity so health service provider finds difficulty in analyzing the data. In database the enormous data has been grouped in terms of clustering. Without getting help from class labels the data can be separated into multiple set of data and can receive data inputs in k means clustering. This research works and implements on the disease data of the patient using k-means clustering. The functionality technique has been contributed by the data extraction approaches to reform the heterogeneous information’s into useful quality information for making decisions. This article educates application and uses of data mining in assist care medical field. Importantly, we learn large dataset clustering on k-means clustering algorithm and produce an improvement to k-means clustering, which needs k or peripheral amount of data which is passed to the dataset. We suggest an algorithm called as G-means, which avail a greedy method to generate preparatory centroids and from that it takes k or peripheral progress among the given data in datasets to make modification in centre points.  Our exploratory outcome which has been used in developing way on the similar data set, displays to us that G-means surpasses k-means in phase of F-scores and entropy. The execution time and coefficient of variance has been executed and it produces best outcomes for G-means clustering.