An SVM-based predictive system for detection and forecasting of unlicensed FM broadcasts

Unlicensed FM broadcasting continues to pose significant risks to national security, communication integrity, and regulatory oversight, particularly in regions where radio remains a vital means of public discourse. This study presents an intelligent, parameter-based system for detecting and forecasting unauthorised FM broadcasts, bypassing the limitations of traditional, content-focused approaches. By integrating Support Vector Machine (SVM) classification with Support Vector Regression (SVR) forecasting, this work introduces a proactive monitoring framework that transitions spectrum governance from reactive detection to predictive enforcement. The system analyses key transmission features—assigned frequencies, band occupancy, and stereo multiplex—to distinguish between licensed and illicit broadcasts, independently of content, language, or programming format. Trained on regulatory data from Nigeria's National Broadcasting Commission, comprising 22,971 raw FM spectrum records collected from Abuja and surrounding states over three years (2021–2023), and pre-processed into 3,169 samples, the model achieved 99.96% accuracy, 100% recall, 100% specificity, and 0% false alarm rate, surpassing existing benchmarks by 0.13%. While this improvement may appear numerically modest, it holds operational significance in large-scale regulatory scenarios where even marginal gains prevent false enforcement actions and reduce resource wastage. The integrated forecasting layer (RMSE=0.042, MAE=0.031) enables pre-emptive identification of interference patterns up to 12 months ahead, demonstrating strong predictive alignment with 2024 interference cases. This work presents a scalable, computationally lightweight, and data-driven solution for spectrum monitoring. It applies beyond Nigeria to spectrum-constrained developing regions, enhancing regulatory capabilities in densely populated environments.