Social Media Fake Profile Detection Using Data Mining Technique

Abstract

In social media, a significant amount of data has been distributed in the entire world with thousands of new users joining social media each day. Social media is a virtual life where malicious users can impact someone’s reputation. Mostly such kind of activity is performed by fake accounts. Thus, identification of fake profiles is necessary and can be done in the early stage of profile building is an essential task for ML. In this paper, the aim is to design a ML model which identifies fake profiles in the early stage and ML based survey on social media has been carried out. Further, the collected literature is categorized according to the used social media datasets and popular areas of employing ML in social media platforms. In this investigation, we have used the Twitter dataset fake profile detection to demonstrate the proposed idea of ML-based fake news detection. The proposed model includes preprocessing to refine the contents and attributes to improve the quality of the dataset and reduce dimensions of the data. The next five popular ML algorithms namely C4.5, Bayes classifier, SVM, ANN, and KNN algorithms are implemented to predict the fake profiles. The evaluation of the system is performed under two scenarios based on training and testing sample ratio of 70-30% and 80-20% and using 4-fold cross-validation. Findings show 80-20% based samples reduce the resource consumption and 70-30% of ratio improves the classification accuracy. Finally, the future extension of the presented work has been discussed.