Simulated Log Data Generation for AI Network Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
AI-based network management systems face challenges in maintaining up-to-date rules due to the need for rich training data, which requires prolonged network operation, leading to outdated rules as network configurations and loads change rapidly.
Innovation Solution
A system and method for generating simulated log data based on new network configurations, allowing for the supplementation of existing log data and enabling the training of AI-learning engines with updated rules without requiring extensive operational time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the AI-rule-generation engine waits for rich training data from actual network operation, then the training data quality improves, but the rules become outdated as network configurations change
Solution Approach 1:
The system performs preliminary action by generating simulated log data in advance for new network configurations before actual operational data is collected. The simulation engine creates artificial training data that reflects the new configuration, allowing the AI rule-generation engine to be trained proactively rather than waiting for sufficient real-world data to accumulate, thus preventing rule outdatedness while maintaining training quality
Solution Approach 2:
The system uses copying by creating simulated log data that replicates the characteristics and patterns of actual network log data. The simulation engine generates artificial data that copies the statistical properties, event types, and relationships found in real network operations, providing high-quality training data without requiring prolonged actual network operation in new configurations
2Quantity of substance
If the communication network operates for a relatively long time to generate rich log data, then the training data becomes more comprehensive, but the rules lag behind current network configurations
Solution Approach 1:
The system generates simulated training data in advance for new network configurations before deploying them to production. By proactively creating comprehensive training datasets through simulation, the AI rule-generation engine can be retrained immediately when configurations change, ensuring rules remain current without requiring lengthy operational periods to accumulate sufficient real data
Solution Approach 2:
The simulation engine creates copies of actual network log data patterns and characteristics, generating artificial datasets that replicate the volume, diversity, and complexity of real operational data. This copying approach provides comprehensive training data instantaneously, eliminating the need to wait for extensive real-world operation while maintaining data quality and variety
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A system, method, and computer program product are provided for training an AI-based network management system, in accordance with one embodiment. In use, log data and first network configuration data are received for a first configuration of a communication network. Additionally, second network configuration data is received for a second configuration of the communication network. Further, simulated log data is produced for the second configuration of the communication network, based on the log data and the second network configuration data.