Synthetic Network Traces for Predictive KPI Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Networks are fragile to conditions and events that have not yet occurred due to the rarity of failures and issues, making it difficult for machine learning models to predict and prevent them effectively, as they require observed patterns to recognize behavioral patterns leading up to such events.
Innovation Solution
A service combines telemetry data from multiple networks into a synthetic input trace and applies it to machine learning models trained on different networks to generate predicted key performance indicators (KPIs), allowing for the identification of abnormal behaviors and potential issues across networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are trained on observed network data to predict failures, then the models can recognize behavioral patterns leading up to failures, but the network remains fragile to conditions and events that have not yet occurred because failures are rare occurrences
Solution Approach 1:
The patent creates synthetic network traces by copying and combining features from multiple real network traces. These synthetic traces replicate the structure and behavior of actual network data while including rare failure patterns that would be insufficient in real observations alone. The synthetic traces are generated by selecting and combining features from multiple source traces to create new training data that preserves important behavioral patterns.
Solution Approach 2:
The patent merges multiple network traces together to create synthetic training data. By combining features from different real traces, the system creates enriched training samples that include rare failure conditions. This merging process allows the machine learning model to learn from patterns that would be too scarce in any single real network's historical data.
2Quantity of substance
If the network collects and stores detailed telemetry data from all networks, then the machine learning models have more training data, but the system complexity increases
Solution Approach 1:
The patent extracts only the necessary features from the full telemetry data. Instead of processing complete network traces, the system identifies and extracts specific feature sets that are most useful for training machine learning models. This extraction process filters out redundant information while preserving the critical patterns needed for prediction.
Solution Approach 2:
The patent segments the network data into distinct feature sets and trace components. By dividing the complex telemetry data into manageable segments, the system can process and combine these segments more efficiently. The segmentation allows for targeted processing of specific network parameters rather than handling the entire data stream as a monolithic complex structure.
Data Source
AI summary
In one embodiment, a service receives telemetry data collected from a plurality of different networks. The service combines the telemetry data into a synthetic input trace. The service inputs the synthetic input trace into a plurality of machine learning models to generate a plurality of predicted key performance indicators (KPIs), each of the models having been trained to assess telemetry data from an associated network in the plurality of different networks and predict a KPI for that network. The service compares the plurality of predicted KPIs to identify one of the plurality of different networks as exhibiting an abnormal behavior.