ML Model Assessment Feedback Loop for 5G Network Automation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 5G network data analytics systems lack effective mechanisms for continuously assessing and improving the accuracy of machine learning models used for network automation, leading to suboptimal performance and decision-making.
Innovation Solution
The proposed solution involves a method and apparatus for providing a machine learning model assessment process, where a provider offers a first ML model to a consumer, collects data from various sources, and based on the assessment, provides a second ML model or a retrained model, incorporating features like model training logical functions and analytics logical functions to monitor accuracy and provide analytics feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a machine learning model is deployed for network automation, then network automation capability is improved, but model accuracy deteriorates over time due to lack of continuous assessment and retraining mechanisms
Solution Approach 1:
The patent implements a feedback mechanism where the provider NWDAF continuously monitors the accuracy of ML models deployed at the consumer NWDAF. The consumer NWDAF sends analytics feedback information and accuracy information back to the provider, enabling continuous assessment and triggering retraining when accuracy deteriorates below thresholds, thus maintaining reliability while preserving automation capability
Solution Approach 2:
The patent establishes preliminary assessment mechanisms by collecting data from multiple data sources (data source NF, DCCF, ADRF, MDAS, UDM) before deploying models and continuously monitoring accuracy. This preliminary and ongoing assessment infrastructure is set up in advance to detect accuracy deterioration and trigger retraining before the model becomes completely unreliable
2Reliability
If ML model assessment and retraining mechanisms are implemented, then model accuracy is improved, but system complexity increases due to multiple data collection functions and coordination
Solution Approach 1:
The patent segments the complex assessment system into distinct functional components: provider NWDAF with MTLF for model training and assessment, consumer NWDAF with AnLF for analytics generation, and specialized data collection functions (DCCF, ADRF, MDAS, UDM). Each component has a specific responsibility, making the overall complex system manageable through clear functional separation
Solution Approach 2:
The patent creates universal interfaces and standardized data collection mechanisms that can serve multiple purposes. The same data collection infrastructure (DCCF, ADRF) serves both initial model training and continuous accuracy assessment. The standardized service operations (Nudm_SDM_Subscribe, Nudr_DM_Subscribe) provide multi-functional data access across different assessment needs
3Measurement precision
If data is continuously collected from multiple sources for model assessment, then model accuracy monitoring is improved, but data collection time and resources increase
Solution Approach 1:
The patent performs preliminary data collection and caching through the ADRF (Analytics Data Repository Function) and DCCF (Data Collection Coordination Function). Data from multiple sources (MDAS, UDM, data source NF) is pre-collected and stored in standardized formats with identifiers (ADRF ID, DataSetTag), enabling rapid retrieval during accuracy assessment without real-time collection delays
Data Source
AI summary
A method of providing a model for analytics of network data and apparatuses for performing the same are provided. An operating method of a provider providing a machine learning (ML) model comprises providing a first ML model to a consumer, assessing the first ML model by collecting data from data sources, and based on an assessment result of the first ML model, providing a second ML model to the consumer.