Proposed Efficient Clustering Technique for Data Mining: The Design and Implementation

Abstract

The large data can be collected from distinguish resources and make a pattern from the collected data. The pattern of the data can be used to create a useful information and for taking decisions in the critical situation. Data mining is a method which is used to extract the important information using some predefined methods or techniques. Clustering is a method of data mining which is used to clusters a group of similar kind of objects or information. There are some clustering techniques such as k-means, k-mode, and k-prototype used to apply in numerical datasets. The proposed work does not require an estimation or value of clusters k for the input purpose. It indicates the outlier to the object which is not satisfy the objective function. To evaluate the performance of the generated clusters using the comparison with enhanced k-means and existing method. The algorithms developed in the present work have been implemented in C# and the output of the original algorithms are obtained using software RapidMiner. The results have been verified on various datasets of different dimensions and sizes ranging from three to fifteen in dimensions and hundreds to thousands in size. The datasets have been obtained from UCI Machine Repository and KEEL Data Repository. The performance evaluation can be measured with existing k-means algorithm and also shows a comparison with k-modes, k-means and K-Prototype algorithms.