FAKE ACCOUNT DETECTION USING MACHINE LEARNING AND DATA SCIENCE

Abstract

Nowadays, online social media platforms play a dominant role in connecting people across the world. The number of users on social networking platforms is increasing rapidly, making communication easier and faster. However, this rapid growth has also created opportunities for malicious activities such as fake identities, spam accounts, and the spread of false information. Recent studies indicate that the number of accounts on social media platforms is significantly higher than the number of actual users, suggesting the widespread presence of fake accounts. Fake accounts are commonly used for activities such as spreading misinformation, posting spam advertisements, and manipulating public opinion. Detecting these accounts has become a major challenge for social media service providers. Traditional detection methods often struggle to accurately differentiate between genuine and fake users, especially with the increasing sophistication of automated account creation techniques. Earlier research utilized machine learning algorithms such as Naïve Bayes, Support Vector Machine (SVM), and Random Forest for fake account detection. However, with evolving patterns of fake account behavior, these approaches have shown limitations in maintaining high detection accuracy. To address this issue, this research proposes an improved detection approach using the Gradient Boosting algorithm combined with Decision Trees. The model focuses on three key behavioral attributes: spam commenting activity, artificial user behavior, and engagement rate. By integrating Machine Learning techniques with Data Science analysis, the proposed system aims to efficiently classify social media accounts as real or fake. Experimental results demonstrate that the proposed approach improves detection accuracy and provides a more reliable solution for identifying fraudulent social media accounts.