Objective: To predict the fraudulent ranking behaviour for mobile apps in which the fraudulent evidences are generated by thew mobile app developers for providing the top ranking for them.
Methods: In mobile app development, the greatest challenge is ranking the mobile app by fraudulent behaviour. The fraudulent behaviour is performed because of the degradation of significant level of the mobile apps. In previous work, to leverage the fraudulent ranking behaviours, Leading Session Methodology based Evidence Aggregation (LSMEA) and Concept Vector based Review Evidence Analysis (CVREA) are developed. These methods consist of ranking based evidences, rating based evidences and review based evidences. Then, these evidence results are aggregated to detect the fraudulent ranking behaviour of mobile apps. In which, rating based evidence is described based on the user rating for corresponding mobile app. User rating is the most significant features for advertising the mobile apps. In previous analysis, user rating is analyzed by using a Gaussian distribution for computing p-value of the statistical hypotheses which is used to define the probability of user rating based on leading sessions.
Findings: The rating based evidences analysis based on the Gaussian distribution suffers from some important limitations. For large dimensions, the total number of parameters is increased quadratically and the manipulation and inversion process of large matrices may become prohibitive. In addition, Gaussian distribution is intrinsically uni-modal. Therefore, a better approximation is not provided for multimodal distributions. Such limitations are removed by introducing Proportional Reversed Hazard Model (PRHM). In this paper, an improvement of finding fraud ranking of mobile apps (IFFR) is proposed by using PRHM and the three evidence outcomes are combined for detecting the fraudulent ranking behaviour for mobile apps.
Applications/Improvements: Mostly, the proposed approach is useful for mobile app markets for developing more number of apps for the specific purpose. Therefore, the accuracy and reputation level are required for further improvement. Thus, the reputation level and accuracy is improved and the fraudulent ranking behaviour of mobile apps is removed by the proposed approach.
Keywords
