Learning Method For Data Clustering

Clustering is the process of grouping a set of objects into one based on their similarity metric. From past many years Clustering has been one of the exploratory research areas for many of the researchers. We have a wide number of clustering algorithms from past research. But none of the algorithms are proven to be work well when we use large set of data items. The greater part of existing clustering algorithms stumbles upon severe scalability and/or accuracy allied problems when used on databases with a huge number of records and/or attributes. In this paper, we propose a hybrid technique, which is based on two concepts namely ID3 decision tree and PFCM. The new technique is able to overcome many of these shortcomings. The primary intension of the research is clustering the data with efficient technique. The key idea is to create an unsupervised decision tree based on the data set which maximizes the difference between the feature-value means of the children nodes. We achieve this by employing ID3 decision tree algorithm. By using only best splits we can reduce the time. To make the clustering efficient we use possibility fuzzy c-means clustering on the results obtained from ID3.This method aims for large scale data. 1