Predilection decoded: Web based spam detection and review analysis for online portals

Abstract

A vast majority of people depend on pre-existing information available on social media to aid them in their decisions. The most common being: Reviews on various products available in the market. With internet services being provided to any and every human being, there are certain drawbacks with such as leaving negative or disingenuous reviews about various products and services offered on internet platforms varying in interests. The classification and determination of such spammers along with the spam content is quite growing topic for analysis and more deep research. A substantial quantity of researches have been carried out regarding this topic, however, the methodologies that have been presented are of high complexity and do not have an easy to use interface for the same. In this research paper, we put forth a simple yet highly effective framework that uses basic algorithms of cosine similarity and sentiment analysis, to implement a web-based model for spam and fake review detection. We segregate the comments as fake, meta-fake and genuine reviews. Sentiment Analysis, Negative Ratio Checking and Cosine Similarity are used for detection of fake reviews and spam content along with other examinations. Incorporating changes based on customer feedback is one of the most important activities carried out by product designers. Spam detection and fake review identification can help an organization analyze, improve and enhance their product based on the suggestions in the real classified reviews given by the customers. If this information is made public by the organization, people can decide whether to buy the product or not based on the real reviews that have been identified by the system.