ABSTRACT

Clustering is tend to be an important issue in data mining applications. Many clustering algorithms are available to cluster datasets that contain either numeric or categorical attributes. The real life database consists of numeric, categorical and mixed type of attributes. It is an essential task to cluster these data sets to extract significant knowledge from the existing database or to obtain statistical information about the database. Clustering large database is a time consuming process. Sampling is a process of obtaining a small set of data from the large database. Applying sampling technique would not cluster all the data points. Labeling non- clustered data point is an issue in data mining process. This paper mainly focuses on clustering mixed data set using modified MARDL (MAximal Resemblance Data Labeling) technique and to allocate each unlabeled data point into the corresponding appropriate cluster based on the novel clustering representative namely, N-Nodeset Importance Representative (NNIR). Accuracy and Error rate are considered as the metrics for evaluating the performance of the existing and proposed algorithm for mixed data set. The experimental result shows that MARDL for mixed data set algorithm performs better than the existing enhanced k-means.