Performance Evaluation of Clustering-Based Classification Algorithms for Detection of Online Spam Reviews

Abstract

The social Web contains enormous content, in which a higher range of online users believes the online reviews for their decision making before any online purchase or using services. The online reviews written are all not true, and some of them are Spam. Data mining classification helps in finding the review as spam or ham. Many text classification algorithms are existing, and it has been proved that these classifiers can be improved when clustering is used with classification to form features. This research work focuses on finding out the right classifier and improving it through clustering. Also, uncover the suitable clustering technique for improving the performance of the classifier. Three classifier algorithms such as Naive Bayes, support vector machine (SVM) and Decision Tree classifiers; clustering algorithms such as K-means, One-Pass and DBScan (Density-based spatial clustering of applications with noise) clustering algorithms are compared and found that the K-means clustering, when used with SVM classifier, outperforms than other combinations. Amazon and Yelp datasets are used for implementation, and the accuracy of SVM classifier has been improved from 89.02% to 90.02% for Amazon dataset and from 86.03% to 88.25% for the Yelp dataset when used with K-means clustering, which is significant than the other combinations compared.