A Comparative Machine Learning Framework for Detecting Fake Accounts on Facebook
Shruti Thapar, Gaurav Kumar Soni, Harshita Kaushik, Reena Singh, Smita Bisht, Sonia Kaur Bansal
Abstract
As social media sites are on the rise, the issue of spam and false accounts has become a major concern of protection, which consequently impacts on user satisfaction and platform reputation. In this study, suggest machine learning model to identify fake accounts in Facebook. The main goal is to create the efficient and correct system able to classify the Facebook accounts either as FAKE ACCOUNT or NOT FAKE ACCOUNT. Research resolves the most important issues in the proliferation of fake accounts by creating an easily comprehensible framework and the use of several machine learning algorithms, such as Logistic Regression, Random Forest, Support Vector machine (SVM) and Naive Bayes. To determine the most effective model, the performance of these algorithms is compared with accuracy and other parameters of performance. According to experimental findings, Logistic Regression, Random Forest, SVM and Naive Bayes achieved an accuracy of 96.8%, 97.2%, 96.2% and 92.3% respectively, thereby illustrating high reliability to identify fraudulent accounts. The findings demonstrate the prospect of machine learning methods in the automated security system improvements and in increasing the reliability of social media.