Sentiment Analysis and Classification on Twitter Spam Account Dataset

Abstract

The amount of people using social media is very large and is increasing day by day. The impact of public figures in social media is quite big. Fake accounts are created in social media platforms and are used for various purposes like inflating the follower list of a particular account. These accounts also called spam accounts usually post spam messages which are used for marketing certain products or spreading particular agendas. Such accounts can be dangerous as they may alter a normal user’s perspective on certain topics. These accounts are used to modify and help in creating a fake sense of popularity which can influence political and social situations. In this project, we try to examine some of the existing methods and approaches for fake Twitter accounts detection. We will make use of a public dataset which contains tweets and account information of both Legitimate accounts as well as spam accounts. We make use of account information to create a classifier which can easily classify whether the given account is a fake account or a legitimate account. We also apply sentiment analysis algorithms on the tweet to find patterns among them. We try to analyse the sentiments behind the tweets of different accounts. Comparing our model with the existing model we will improve the features present in our model. In the process of building a better model, we try to reduce overfitting. The final result is an optimum classifier, which can be used to separate a fake account from a list of legitimate accounts.