International Journal of Computer Applications |
Foundation of Computer Science (FCS), NY, USA |
Volume 185 - Number 9 |
Year of Publication: 2023 |
Authors: Srinivasa Suresh Sikhakolli, Asha Kiran Sikhakolli |
![]() |
Srinivasa Suresh Sikhakolli, Asha Kiran Sikhakolli . Effective Purity Method for Measuring the Clustering Accuracy and its Illustration. International Journal of Computer Applications. 185, 9 ( May 2023), 28-33. DOI=10.5120/ijca2023922752
Clustering is one of the commonly used model in business and scientific applications. Often, data science specialists and researchers apply clustering techniques for classification and optimization. Measuring the clustering accuracy is one of the key parameter. There are several extrinsic measures exists for measuring clustering quality. One of them is Purity. It indicates the level of homogeneity of the clusters. Purity computes the sum of frequencies of the dominant class in each cluster and then divides the sum by total number of records. In the existing purity method, total number of clusters is not taken into consideration. According to the researcher, number of clusters have significant effect on overall cluster quality. In this paper, the researcher proposed an algorithm with few changes to the existing purity method. The proposed algorithm is applied on machine learning data sets taken from UCI machine learning repository. Further, significant improvement in purity computation is observed when applied using FCM and K-means clustering. This paper explains proposed algorithm artificial illustration, results & analysis and comparative analysis between proposed purity method an existing purity method.