Classification of Fake, Bot, and Real Accounts on Instagram Using Machine Learning

ABSTRACT

Instagram is a social media platform that allows users to share content such as photos and videos. Fake and bot account problems constitute a significant obstacle to social networking. Since fake and bot accounts have purposes such as increasing the number of followers, creating a perception by using misinformation, deceiving people, detecting these fake and bot accounts plays an essential role in creating a secure social network. Fake account detection is beneficial to keeping people safe from misinformation and malicious profiles on Instagram, ensuring customers' safe accounts, and preventing fraud. From this point, we aim to classify Instagram user profiles into fake, bot, and real accounts with classification algorithms. Additionally, we present a publicly available dataset for the fake, bot, and real accounts detection on Instagram. For data collection, real accounts were determined from our circle of friends, fake accounts were accessed by manual scanning from Instagram, and bot accounts were accessed by purchasing from bot account websites and mobile applications. These accounts' features were collected via web scraping. We use the seven classifiers to train classification models in fake, bot, and real profile detection. Our results show that the Random Forest gives the highest prediction accuracy with 90.2%.