An Adaptive Sensing and Regulatory Model for Illegal FM Interference Mitigation using SVM

Illegal FM broadcasting poses a persistent challenge for regulatory agencies, particularly in developing nations such as Nigeria, where limited monitoring infrastructure and enforcement capacity enable unauthorised broadcasters to exploit spectrum loopholes. This study proposes an adaptive sensing-and-regulatory model using a Support Vector Machine (SVM) to detect illegal FM interference. The model leverages regulatory signal parameters—assigned frequency, band occupancy, and stereo multiplexed signal (MPX)—to classify transmissions as licensed or unlicensed. Three distinct contributions advance the field: (i) regulatory-aware feature engineering using assigned frequency, band occupancy, and stereo multiplexed (MPX) signal parameters aligned with actual enforcement protocols; (ii) adaptive cross-regional validation demonstrating generalizability across Nigerian metropolitan areas; and (iii) integration of real-time sensing capabilities within a computationally efficient framework suitable for resource-constrained regulatory environments. The model processes spectrum data to distinguish between legal and unauthorised transmissions. A comprehensive dataset sourced from a metropolitan FM environment was pre-processed to eliminate redundancy and trained using randomised cross-validation with optimal kernel tuning. The developed model achieved an accuracy of 99.94% on hold-out test data. To address concerns about overfitting, rigorous validation strategies were employed, including stratified train-test splits, RandomisedSearchCV with 5-fold cross-validation, and successful deployment on geographically distinct, unseen datasets from Lagos, Kano, and Sokoto, which confirmed model robustness and outperformed comparable existing approaches. Comprehensive evaluation metrics—including Precision (99.41%), recall (99.65%), F1-score (99.53%), and a false-alarm rate of 0.06% validate the model's suitability for regulatory enforcement. By integrating adaptive sensing with intelligent classification, the model provides a scalable regulatory tool that supports real-time enforcement decisions and proactive spectrum governance.