Parallel Data Analysis Modules for Mobile Access Network Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data analysis methods for mobile access networks are inadequate in processing speed, particularly due to the short validity period of wireless air interface data, which poses a high requirement for rapid data analysis in these networks.
Innovation Solution
A data analysis apparatus comprising multiple modules that work in parallel to quickly analyze data from network nodes in mobile access networks, generating feature vectors, creating model instances, and adjusting configuration parameters based on prediction results, thereby optimizing network performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data analysis is performed using traditional sequential methods, then processing accuracy can be maintained, but processing speed is insufficient to meet the short validity period of wireless air interface data
Solution Approach 1:
The patent segments the data analysis system into multiple independent modules (feature extraction module, model instance creation module, prediction module, parameter adjustment module) that operate in parallel. Each module processes specific aspects of data analysis independently, allowing simultaneous execution of multiple analysis tasks which significantly improves processing speed while maintaining manageable complexity through modular design
Solution Approach 2:
The patent transitions from sequential single-threaded processing to parallel multi-threaded processing by adding a temporal dimension to the execution model. Multiple model instances process different feature vectors simultaneously across different time slices, effectively increasing processing throughput without proportionally increasing physical hardware resources
2Reliability
If model training is performed continuously to maintain accuracy, then prediction accuracy is improved, but processing time increases due to the overhead of retraining
Solution Approach 1:
The system performs preliminary model training offline to create a trained model, then uses this pre-trained model for rapid online predictions. Model updates are performed periodically or event-driven rather than continuously, allowing the system to maintain good prediction accuracy while avoiding the time overhead of constant retraining. Feature extraction and prediction can proceed using the pre-trained model structure
Solution Approach 2:
The patent creates multiple copies (instances) of the trained model that can process different feature vectors simultaneously. Instead of retraining a single model continuously, multiple model instances are deployed to handle different data streams or feature sets in parallel, maintaining prediction accuracy across multiple concurrent operations without requiring repeated training cycles
Data Source
Figure 1(a)~2
Figure 3
Figure 4
AI summary
This application provides a data analysis apparatus, system, and method. The data analysis apparatus may collect data of a network node in a mobile access network, and then perform data analysis and adjust a configuration parameter of the network node, to implement targeted data analysis and parameter adjustment for the network node, so that a speed of data analysis and configuration parameter adjustment can be improved. In addition, modules of the data analysis apparatus may run in parallel, so that the speed of data analysis and configuration parameter adjustment can be further improved.