RAN Antenna Issue Detection Using Uplink PRB Interference

The patent transforms the static threshold parameter into dynamic thresholds that adapt based on historical data, environmental conditions, and learned patterns from machine learning models. This allows the system to maintain ease of operation while significantly improving reliability by reducing false positives and enabling faster identification of actual antenna issues.

The patent replaces the simple mechanical threshold comparison system with an intelligent system using machine learning models (neural networks, decision trees, random forests) that automatically analyze KPI data patterns. This substitution maintains operational simplicity while dramatically improving detection reliability by learning from historical data and adapting to changing conditions.

The patent divides the complex analysis task into multiple specialized machine learning models, each trained to detect specific antenna issue types (alignment problems, cable issues, connector problems). This segmentation allows each model to focus on particular patterns, improving measurement precision while organizing complexity into manageable, modular components that can be independently maintained and updated.

The patent introduces data preprocessing and feature extraction layers as intermediaries between raw KPI data and the machine learning models. These intermediaries transform complex raw data into standardized features, reducing the complexity burden on individual models while maintaining high measurement precision through systematic data preparation.

The patent performs preliminary analysis using multiple machine learning models to pre-filter and prioritize alerts before human investigation. The system automatically scores and ranks potential issues based on likelihood and severity, so technicians only need to investigate high-confidence cases. This preliminary action maintains reliability by ensuring actual issues are identified while dramatically improving productivity by eliminating the need to manually investigate false positives.

The patent implements feedback loops where investigation outcomes (confirmed issues vs. false positives) are fed back into the machine learning models to continuously improve their accuracy. This feedback mechanism increases reliability over time by learning from real-world outcomes, while maintaining productivity by reducing the volume of false alerts requiring manual investigation.