Fake Account Detection using Machine Learning and Data Science

Nowadays, Online Social Media is dominating the world in several ways. Day by day the number of users using social media is increasing drastically. The main advantage of online social media is that we can connect to people easily and communicate with them in a better way. This provided a new way of a potential attack, such as fake identity, false information, etc. A recent survey suggest that the number of accounts present in the social media is much greater than the users using it. This suggest that fake accounts have been increased in the recent years. Online social media providers face difficulty in identifying these fake accounts. The need for identifying these fake accounts is that social media is flooded with false information, advertisements, etc. Traditional methods cannot distinguish between real and fake accounts efficiently. Improvement in fake account creation made the previous works outdated. The new models created used different approaches such as automatic posts or comments, spreading false information or spam with advertisements to identify fake accounts. Due to the increase in the creation of the fake accounts different algorithms with different attributes are use. Previously use algorithms like naïve bayes, support vector machine, random forest has become inefficient in finding the fake accounts. In this research, we came up with an innovative method to identify fake accounts. We used gradient boosting algorithm with decision tree containing three attributes. Those attributes are spam commenting, artificial activity and engagement rate. We combined Machine learning and Data Science to accurately predict fake accounts.