An Efficient Automatic Classification Approach To Detect Average Over Popular Item Attack In Recommender System

Abstract

newline Social media is used widely nowadays. This platform is widely used for promoting products to increase sales. The problem arises when the vendor uses fake profiles for promoting products. Detecting these fake profiles is the need of the hour. Many different mechanisms through content and collaborative filtering are proposed. Content-based filtering is based upon the content of the product. Collaborative filtering on the other hand is based upon the content of the product as well as user preferences. The proposed approach is based upon the fake movie promotion through social media and detecting such fake promotions. This dataset contains noise in the form of missing values. In addition, string literals are present. These literals must be converted into nominal form to enhance the classification accuracy. To achieve the entire objective of fake promotion, a MovieLens dataset is used. This dataset is available on the MovieLens website. The first phase in the proposed system is pre processing. The pre-processing mechanism includes null value handling and a nominal conversion mechanism. After this phase feature extraction takes place. Social media provide a platform for different users to interact with each other. This platform is not only attracting people but also business owners for promoting the products. The most common use of social media platforms is to promote movies online. For this purpose, fake profiles are created by the promoters. Detecting such fake profiles to uncover actual ratings for the movie is the prime objective of this research. To detect the fake profile phases are followed. These phases include pre processing, segmentation, and classification. Pre-processing mechanism eliminates the noise if any from the movie dataset.