NWDAF Prediction Optimization Through Time-Window Algorithm Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3GPP Network Data Analytics Function (NWDAF) techniques use a single forecasting algorithm for all workloads, leading to suboptimal tradeoffs between accuracy, latency, and resource consumption, and lack efficient handling of high volumes of dynamically requested data from consumer devices.
Innovation Solution
The NWDAF employs a smart prediction optimization approach that deconstructs subscription requests into analysis tasks, groups them by common attributes, selects appropriate forecasting algorithms based on prediction time windows, and tunes input data window sizes using machine learning to optimize accuracy, latency, and resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single forecasting algorithm is used for all workloads in NWDAF, then device complexity is reduced, but manufacturing precision (prediction accuracy) deteriorates
Solution Approach 1:
The patent segments the workload into different categories (e.g., immediate, short-term, long-term predictions) and assigns different forecasting algorithms to each category. This segmentation allows the system to use simple algorithms for straightforward cases while applying more sophisticated algorithms only where needed, thereby maintaining prediction accuracy without uniformly increasing complexity across all workloads.
Solution Approach 2:
The patent applies local quality by selecting different algorithm complexities for different prediction scenarios. Instead of using a uniformly complex algorithm throughout, the system applies appropriate algorithmic sophistication locally to each workload type, matching the computational resources to the specific requirements of each prediction task.
2Ease of operation
If a single forecasting algorithm is used for all workloads, then ease of operation is improved, but productivity deteriorates
Solution Approach 1:
The patent introduces dynamic algorithm selection based on workload characteristics and prediction time windows. The system automatically adapts the algorithm choice to match the specific requirements of each prediction task, improving processing efficiency without requiring manual intervention. This dynamic adaptation maintains ease of operation while significantly enhancing productivity.
Solution Approach 2:
The patent changes operational parameters (algorithm selection) based on the characteristics of the workload and prediction time window. By dynamically adjusting which algorithm is applied based on parameters such as prediction duration and data patterns, the system improves processing efficiency while maintaining simple operation through automated parameter selection.
3Device complexity
If conventional NWDAF techniques are used, then device complexity is reduced, but loss of time (latency) worsens
Solution Approach 1:
The patent segments predictions by time window characteristics (immediate, short-term, long-term) and applies appropriately optimized algorithms to each segment. This segmentation enables the system to use fast, lightweight algorithms for immediate predictions while allocating more computational resources to long-term predictions, thereby reducing overall latency without requiring complete system redesign.
Solution Approach 2:
The patent performs preliminary classification of workloads based on prediction time windows and characteristics before applying forecasting algorithms. This preliminary action allows the system to prepare and select the most appropriate algorithm in advance, reducing processing latency while maintaining manageable system complexity through structured workflow organization.
4Ease of operation
If conventional NWDAF techniques are used, then ease of operation is improved, but resource consumption worsens
Solution Approach 1:
The patent dynamically changes operational parameters (algorithm selection, input data window size) based on workload characteristics and prediction requirements. This parameter optimization reduces computational resource consumption by applying appropriately scaled algorithms to each task, avoiding the waste of using overly sophisticated algorithms for simple predictions while maintaining simple operation through automated parameter management.
Data Source
AI summary
Systems and methods are provided for executing NWDAF uses cases with a more optimal tradeoff between accuracy, latency, and resource consumption, to thereby improve upon network performance or services. For example, the systems and methods can provide smart NWDAF prediction optimization that includes grouping 3GPP subscription requests based on common attributes of analysis tasks, creating a single analysis prediction workload for each group of the analysis tasks, and selecting a forecasting algorithm for executing each single analysis prediction workload based on a prediction time window classification.