AN EFFECTIVE FRAMEWORK FOR DATA CLUSTERING USING IMPROVED K-MEANS APPROACH
Sakshi Siva Ramakrishna · International Journal of Advanced Research in Computer Science · 2018
Data clustering refers to the partition of a dataset into homogeneous subsets where each subset is dissimilar to the rest of the subsets. K-means is a familiar approach for data clustering particularly when all the attributes of the data objects are of numeric type. Though the k-means approach is popular and efficient it is susceptible to misclassify the data due to the noise and outliers that are common in datasets. The aim of this paper is to study the strategies available to overcome the problems like high dimensionality, redundancy, noise and outliers while implementing the k-means algorithm and to propose a better approach to deal with the problem. An iterative attribute reduction procedure based on correlations among attributes was proposed to cluster the given dataset using k-means algorithm in an improved manner. The standard dataset “Iris” was used to test the proposed methodology. The obtained results are reasonably better.