Real-Time Twitter Spam Detection: A Comparative Study of Machine Learning Models

In today’s world, the rise of social networking apps like Facebook, Instagram, WhatsApp, Twitter, Snapchat, and YouTube has largely replaced in-person interactions, conversations, and information sharing. Spamming through messages has been rising in this segment, where people find misleading messages, which could overall affect the performance of trust in social media. Spammers are using social networking applications to send spam messages. Twitter, which has become one of the critical sources of communication and news propagation, is becoming a base for spammers to post misleading and deceptive content to distract users. Twitter spam frequently includes false information, such as “free coupons” and “weight loss adverts”, to entice users into casting several votes for the same item. Opening unnecessary links and sending unwanted answers or remarks can offer spammers or hackers access to a person’s phone or another device. This will increase the security risks for Twitter as well as the users using Twitter. Due to this, users often fall victim to spam scams through spam messages. To overcome this problem, researchers came up with different techniques to enhance the spam detection performance. So here, we are employing various NLP techniques, like tokenize the lines, stemming, lemmatization, and vectorization methods like Bag-of-Words, Count Vectorizer, and TF-IDF, to preprocess the data and extract features, which are then used for classification through Machine Learning (ML) algorithms. We would fetch the best algorithm for detecting spam messages on Twitter and classifying them into spam and non-spam tweets or spam and non-spam accounts.