Detection of fake identities on twitter using supervised machine learning

Abstract

Social media has changed the way people get their news. Once people used to buy newspapers to get their news but now everything is online. It has changed the dimension of how we receive news altogether. With a growing e ect of social media, it brought us the good, bad and ugly as social media is lled with spams and hate speech along with fake news. One of the crucial problems of social media is fake accounts.We planned to get rid of all fake accounts using machine learning speci cally Arti cial Neural Network model. Our purpose was to lter out fake accounts from all the accounts existing on social media. There has been a lot of work on this subject, though no permanent solution could be found. We have collected data from many sources and used around four classi ers to compare and determine which is the best classi er for our paper. We have used numeric attributes from twitter accounts and based on these attributes we were able to nd out fake accounts. We gave more priority to Arti cial Neural Network as we can give di erent weights to di erent attributes and get a more accurate result. Also, we are using K-Nearest Neighbor, Random Forest, Support Vector Machine and Neural Networks to compare between the algorithms.Twitter is known for their fake account problem as the user base of Twitter has just grown with time so has the fake users. So, our paper is based on how we can detect this fake account and bots along with making Twitter more safer for its users.