Mobile Network Sleeping Cell Detection Using Aggregated KPIs

The patent replaces manual mechanical checking with an automated electronic system that collects KPI data from the network and uses machine learning algorithms to detect sleeping cells. The system automatically analyzes connection establishment failures, message transmission issues, and other network parameters to identify sleeping cells without human intervention, thereby eliminating time consumption and labor costs while maintaining high detection precision.

The patent introduces an intermediary detection system that acts as a mediator between the network operations and the operator. This system collects KPI data from various network components, processes it through machine learning models, and generates alerts about sleeping cells. The intermediary system filters and analyzes large volumes of network data, presenting only relevant information to operators, thus improving detection efficiency without requiring manual checking of all network elements.

The patent implements a universal detection system that can monitor and analyze multiple cells simultaneously using the same KPI collection and machine learning framework. The system is designed to handle variable numbers of cells by dynamically adjusting its analysis scope, allowing it to scale from small to large networks without requiring different detection methodologies. This multi-functional approach maintains consistent detection capability across networks of any size.

The patent segments the detection process into modular components: KPI data collection from individual cells, aggregation of data across multiple cells, machine learning analysis, and alert generation. This segmentation allows the system to process large numbers of cells by dividing the workload into manageable units that can be handled independently and in parallel, thereby reducing the perceived complexity even as network coverage expands.

The patent implements a self-service detection system where the network infrastructure itself provides the data needed for detection through standard KPI collection mechanisms. The machine learning models are trained on historical network data and automatically adapt to changing network conditions without requiring manual reconfiguration. The system serves itself by utilizing existing network telemetry infrastructure, thereby improving detection efficiency without proportionally increasing system complexity.

The patent leverages parameter changes in KPI values over time to detect sleeping cells. The machine learning system monitors temporal variations in connection establishment rates, message transmission success rates, and other KPI parameters. By focusing on changes in these parameters rather than absolute values, the system achieves high detection efficiency using relatively simple computational approaches, avoiding the need for extremely complex analysis systems.