Combating Terrorism: Detecting and Analyzing Extremist Content on X (Formerly Twitter)

This study investigates the efficacy of machine learning algorithms in detecting and evaluating extremist content on social media platforms, and we employ X, formerly known as Twitter. We therefore utilize Kaggle's dataset that contains 83,000 entries from both extremist and non-extremist posts to compare the performance of four machine learning models, namely Support Vector Machine (SVM), Random Forest, Gradient Boosting, and Naive Bayes. Moreover, we examine a Stacking Classifier that leverages the advantages of all four models. The models were assessed using accuracy, precision, recall, F1 Score, ROC-AUC Score, training duration, and prediction time. The findings of our study revealed that the Support Vector Machine (SVM) algorithm was the most accurate and resilient, whereas Naive Bayes showcased remarkable computing efficiency, rendering it well-suited for real-time detection. The Random Forest and Gradient Boosting algorithms offered a well-balanced approach. Notably, the novel Stacking Classifier outperformed individual models, demonstrating the strength of a hybrid methodology. Hence our study presents a thorough structure for improving the identification of extremist material on social media, thereby providing essential knowledge for creating efficient, immediate detection systems to combat online extremism successfully.