Spam Detection on Indonesian Beauty Product Review

Abstract

A product review is one of the most important sources of information that can help consumers to find the most suitable products for their needs. However, there is a chance a reviewer has other intentions than providing an honest review, including advertising the brand or other brands. A review that does not contain any information related to the product’s aspects/features could be considered spam. This paper presents our work on spam review detection, specifically in the domain of beauty products. We used SVM and Logistic Regression classifier and the following features: the review sentiments, product-related features, and review-centric features extracted from the reviews. We classified the beauty product review texts as spam and non-spam reviews. The experimental result showed that the best accuracy percentage was 81%, obtained when we used the sentiments and review-centric features with the SVM algorithm.