Vol. 3, Issue 12, Part H (2017)

Abstract
Data mining can be termed as a process of extracting patterns (knowledge) and modelling query from data which is stored in database. Classification is one among of its concept and techniques. K-means is very popular clustering algorithm used. And for classification there are many algorithms provided. This research article is proposing a novel hybrid mining approach by using modified K-Means and Support vector machine algorithm. Modified K-Means utilized here for making the clusters from given dataset and SVM is utilized for classification (on clustered dataset obtained from modified K-means clustering). Experiments are performed over different datasets which are taken from UCI Irvin. Datasets which are used for comparing clustering algorithm are provided in Table 1 along with their details. Evaluations are done on different datasets of following parameters: Accuracy obtained from new algorithm using confusing matrix which is being created for every dataset and time taken to execute the task. Experimental results show that the proposed algorithm has better accuracy than the traditional algorithms. Results gives better and modified classification of clustered data. Additionally, proposed algorithms provide better clustering result than the existing algorithms.