Improvised Spam Detection in Twitter Data Using Lightweight Detectors and Classifiers

Receiving spam messages is one of the most serious issues in social media, especially in Twitter, which is a widely used platform to reflect the opinions and emotions of an individual publicly as well as focused to a specific group of members with similar thoughts or discussion topic. In such focused discussion groups, getting spam message through social media sites is the most annoying issue. In this paper, a system is developed to detect spam tweets by using four lightweight detectors, namely blacklist domain detector, near duplicate detector, reliable ham detector, and multiclass detector. The detected tweets are then classified using ensemble classifiers such as naïve Bayes, logistic regression, and random forest. Voting method is applied to decide the labels for the tweets obtained after classification process. The proposed system has achieved an accuracy of 79% to detect spam tweets with the help of naïve Bayes classifier method and the value seems to be optimizing further with the availability of more sample data.