Analysis of Ensemble Learning Models for Identifying Spam over Social Networks using Recursive Feature Elimination

摘要

Social Networking platforms are regarded as being the reliable and valued communication medium for transferring information and communicating, used by the millions throughout the world. Users’ reliance on these social networking sites is growing to seek perspectives, updates, alerts, news etc. While it is evident that the online social networks have become a way for information sharing, at the same instant they have rapidly become a medium for spreading misinformation, rumors, unsolicited messages, propaganda, fake news, and so on. It can indeed be said that a social networking platform consists of the two types of users, namely Spammers and Non-Spammers. Spammers typically spread misinformation or share undesirable content on social networking websites, out of malicious intents. In this work, a model is proposed to identify Spammers in Twitter network. This work is based on the user behavior-based and content-basedfeatures like Hashtags, URLs, Mentions, Replies, and Retweets. In this work, the Recursive Feature Elimination is used along with Support Vector Machine, Random Forest, Logistic Regression, Adaptive Boosting, and XGBoost. For data pre-processing, Weka is used, and implemented the five classifiers mentioned with Recursive Feature Elimination using sklearn in Python. Performance measures such as TP Rate, FP Rate, Precision, F-Measure and Accuracy are used for evaluating the performance of the proposed model