Detection of Fraudalent Behaviours using Graph Cluster Analysis on Unsupervised Datasets

The widespread use of mobile Application rulers has resulted in fraudulent or misleading practises aimed at boosting the visibility of applications in the top suggestion list. Indeed, fake application developers are constantly using fraudulent ratings and fake feedback to raise their app sales and commit fraud. Although the value of preventing these forged activities is widely acknowledged, there is a lack of knowledge and research in this field. In this paper, we suggest using the Aggregation Algorithm to discover rating fraud for mobile apps, a better way to test and analyse the playstore datasets. In particular, we propose that through analysing datasets using an unsupervised learning method and clustering, we can accurately locate rating fraud. Furthermore, it uses a supervised learning strategy. By using statistical hypotheses experiments to model the agglomerative behaviours of apps, We look at five different kinds of evidence: ranking-based proof, number of installations evidences, ratingbased evidences, review-based evidences, and App description mining evidences. Finally, after combining all of the evidences, a graphical pattern is created on the imported data, which is used for fraud detection discovery.