Semi Supervised PU learning method for deceptive detection using MKNN

Abstract

Currently, in the era of E-Commerce a huge number of user opinion, user reviews and feedbacks are posted on the portal by thousands of users for any services and products. Reviews posted by the users are important source of information both end of the business. Mostly in online selling customers are rely on reviews before ordering the services online. Unfortunately, there is an problem of deceptive opinions, that is, not by real users. Deceptive reviews are aimed to improve the ratings of low quality products or services (Positive Reviews) or they are aimed to downgrade the high quality services of other businesses (negative reviews). For our research in this research paper we have focused on the identification of every type of deceptive reviews, either it is positive or it is negative. Because of the scarcity of samples of deceptive reviews, we propose to present the problem of the “detection of deceptive opinions
employing PU-learning”. PU-learning (Positive Unlabelled Learning) is a semi-supervised technique that is used to build a binary classifier. Mostly it builds classifier on the basis of positive (i.e., deceptive opinions) and unlabelled examples only. Concretely, we propose a novel approach for detection of deceptive spam using PU Learning and Modified KNN algorithm