Data Mining and Knowledge Engineering

J. I. Sheeba ¹, K. Vivekanandan ¹

Affiliations
1 Department of Computer Science and Engineering, Pondicherry Engineering College, Puducherry-605014, IN

Abstract

Clustering is a technique of automatically grouping similar data into clusters. A large diversity of similarity measures distance functions such as Euclidean distance, Jaccard distance, Pearson Correlation distance, Cosine similarity and Kullback-Leibler Divergence have been implemented for clustering. Fuzzy C means algorithm is implemented for assigning membership to each word point in the cluster. In the same way it is calculated to each cluster center from the origin of remote region between the cluster center and the word point in this process. This proposed framework is used to validate the five similarity measure functions with Fuzzy C means clustering algorithm for finding the effectiveness. To estimate the optimal number of clusters, by implementing the validity measures like purity and entropy. Finally the results are compared five similarity measure functions with Fuzzy C Means clustering algorithm. Euclidean similarity measure function provides better and accurate results as compared to the other distance functions.

Keywords

Clustering, Euclidean Distance, Fuzzy C Means Algorithm, Similarity Measure.

Full Text

     

A. Habiba ¹, J. I. Sheeba ²

Affiliations
1 Department of Computer Science and Engineering, Alpha College of Engineering, Puducherry, IN
2 Department of Computer Science and Engineering, Pondicherry Engineering College, Puducherry, IN

Abstract

Text mining is the significant aspects of study and research motivated by the remarkable growth of the social web. Cyber bully is a major problem that occurs in the online communications. Nowadays, methods for automatic thoughts mining on online data are becoming increasingly important. In this proposal, we studied about the detection of the general cyber bully polarity. A thorough analysis has been made for sentiment (i.e.) opinion mining but the cyber bullies which harass and threatens the online social victims research work has not been done as familiar. The aim of this paper is to extract cyber bully polarity from blog messages with methods of pre-processing, frequency measure and Classifier algorithm. The novelty of this paper arise from treating user generated content on blogs as dynamically evolving linked documents that vary by thoughts, content and emotions.

Keywords

Text Mining, Cyber Bully, Classification, Classifier Algorithm, Abusive Detection.

Full Text