Twitter bot detection using supervised machine learning

Abstract

In the world of Internet and social media, there are about 3.8 billion active social media users and 4.5 billion people accessing the internet daily. Every year there is a 9% growth in the number of users and half of the internet traffic consists of mostly bots. Bots are mainly categorized into two categories: good and bad bots; good bots consist of web crawlers and chat bots whereas bad bots consist of malicious bots which make up 20% of the traffic, the reason they are not good is that they are used for nefarious purposes, they can mimic human behavior, they can impersonate legal traffic, attack IoT devices and exploit their performance. Among all these concerns, the primary concern is for social media users as they represent a large group of active users on the internet, they are more vulnerable to breach of data, change in opinion based on data. Detection of such bots is crucial to prevent further mishaps. We use supervised Machine learning techniques in this paper such as Decision tree, K nearest neighbors, Logistic regression, and Naïve Bayes to calculate their accuracies and compare it with our classifier which uses Bag of bots' word model to detect Twitter bots from a given training data set.