Concurrent Machine Learning Models for RAN-Aware Performance Evaluation

The patent introduces a network evaluation system as an intermediary component that collects network performance data from RANs and core networks, processes this information through machine learning models, and delivers curated network context to mobile devices. This mediator enables devices to make informed connection decisions without directly exposing them to complex network infrastructure details, thereby resolving the contradiction between providing comprehensive network information and maintaining device simplicity.

The system performs preliminary actions by proactively collecting, processing, and storing network performance data before devices need to make connection decisions. The machine learning models pre-process this data to generate predictive insights about network conditions, allowing devices to access ready-to-use network context information without real-time computational overhead, thus improving both information availability and connection decision quality.

The network evaluation system serves as a centralized intermediary that handles all complex data collection, storage, processing, and analysis tasks. By isolating complexity in this dedicated mediator component, the patent allows mobile devices to receive simplified, pre-processed network context information without needing to implement complex data handling capabilities themselves, thus improving connection decision quality while limiting the spread of system complexity.

The patent replaces complex mechanical data processing systems with machine learning models that automatically learn patterns from network data and generate predictive insights. These ML models substitute for complex rule-based processing mechanisms, enabling the system to handle comprehensive network information more efficiently and generate actionable recommendations without requiring overly complex decision-making infrastructure at device level.

The patent extracts only the most critical and relevant network performance indicators from the vast amount of available network data. By identifying and selecting key metrics that directly impact connection decisions, the system transmits minimal necessary information to devices, reducing data transmission overhead and latency while still enabling optimized connection choices. This selective extraction approach resolves the contradiction between comprehensive data sharing and transmission efficiency.

The system performs preliminary processing of network data in advance, pre-computing insights and predictions that will be needed for connection decisions. By preparing and caching this processed information beforehand, the system eliminates the need for time-consuming real-time data processing during connection decisions, thereby reducing transmission time and improving responsiveness while still providing comprehensive optimization capabilities.

The patent implements preliminary action by continuously collecting and pre-processing network performance data in advance of connection decisions. Machine learning models perform real-time analysis of historical data to generate predictive insights about future network conditions, allowing the system to provide accurate performance predictions without requiring computationally intensive real-time calculations during actual connection decisions, thus reducing processing time while maintaining measurement precision.

The system replaces complex real-time computational analysis with machine learning models that have been pre-trained on historical network data. These models substitute for time-consuming mechanical calculations by using learned patterns and relationships to quickly predict network performance outcomes, thereby achieving high measurement precision with significantly reduced processing time and computational resource requirements.