Advances in Mathematical Models for AI-Based News Analytics

Abstract

The exponential growth of digital news sources presents a critical challenge in efficiently processing and analyzing vast datasets to derive actionable insights. This paper introduces a GPT-based news analytics system that addresses this issue using advanced mathematical modeling and AI techniques. Over a 405-day period, the system processed 1,033,864 news articles, categorizing 90.67% into 202 subcategories across 11 main categories. The system achieved an average precision of 0.924, recall of 0.920, and F1-score of 0.921 in event correlation analysis and demonstrated a fast average execution time of 21.38 s per query, enabling near-real time analysis. The system critically analyzes semantic relationships between events, allowing for robust event correlation analysis, with precision and recall reaching up to 1.000 for specific pairs such as “UFO” and “Cyber”. Using dimensional augmentation, probabilistic feature extraction, and a semantic knowledge graph, the system provides robust event relationships for modeling unstructured news reports. Additionally, the integration of spectral residual and convolutional neural networks helps to identify anomalies in time-series news data with 85% sensitivity. Unlike existing solutions reported in the literature, the proposed system introduces a unified mathematical framework for large-scale news analytics, seamlessly integrating advanced methods such as large language models, knowledge graphs, anomaly detection, and event correlation to deliver fast and efficient performance. This scientifically novel and scalable framework offers a transformative approach to solving the pressing problem of news analytics, offering significant value to researchers, policymakers, and media analysts.

Keywords

• GPT-based news analytics
• knowledge graph
• event correlation analysis
• anomaly detection
• spectral residual
• convolutional neural network