Hybrid Lexicon-Driven News Threat Detection Using Random Forest and XGBoost Models

Abstract

The rapid growth of online news platforms and social media has made it easier for information to spread quickly, but it has also increased the circulation of content that may create fear, misinformation, or potential security concerns. Detecting threat-related news at an early stage is essential for supporting public safety and informed decision-making. This study presents a hybrid lexicon-driven news threat detection framework that combines the NRC Emotion Lexicon with advanced machine learning models, namely Random Forest and XGBoost. Initially, news articles undergo preprocessing steps such as text cleaning, tokenization, and stop-word removal to improve data quality. Emotional features are then extracted using the NRC Lexicon and integrated with textual features to create an informative dataset for classification. The processed data is used to train and evaluate both machine learning models using performance measures including accuracy, precision, recall, and F1-score. Experimental results indicate that the proposed hybrid approach effectively identifies threatrelated news, with the Random Forest model providing slightly better classification performance than XGBoost. The combination of emotion-based lexical analysis and ensemble learning enhances prediction accuracy, making the proposed framework a practical and reliable solution for intelligent news threat detection in real-world applications.