Automated Event Prioritization via Co-occurrence Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network Operations Control teams face overwhelming volumes of network events, making it difficult to manually identify and prioritize critical events in a timely manner, as problems in one network component can propagate and cause larger issues, and existing solutions depend on network topology, which can change over time.
Innovation Solution
An automated method and system that creates tuples from events with physical and logical attributes, uses binarized co-occurrence matrices to establish parent-child relations, and applies a heuristic function to identify high-priority events, independent of network topology, by dividing events into samples and calculating probabilistic scores to quantify these relations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual analysis and prioritization of events is performed, then network operators can identify critical events, but the process becomes extensive and time-consuming
Solution Approach 1:
The patent replaces manual mechanical analysis with an automated machine learning system. The ML model automatically processes network events, predicts severity levels, and prioritizes them based on learned patterns from historical data, eliminating the need for manual analysis while maintaining accurate event prioritization.
Solution Approach 2:
The system performs self-service by automatically learning from historical network events and building its own prioritization model. The ML algorithm continuously improves its accuracy by processing new event data without requiring human intervention in the prioritization decision-making process.
2Quantity of substance
If existing event management solutions are used, then events can be classified by severity, but the volume of critical events remains too large to address adequately
Solution Approach 1:
The patent extracts only the most critical events from the entire event stream using ML-based severity prediction. By filtering and extracting only the top priority events that truly require immediate attention, the system reduces the workload for network operators while ensuring that the most important issues are addressed.
Solution Approach 2:
The system changes the parameter of event severity assessment from static classification to dynamic prediction based on ML models. The model continuously refines its severity assessments by learning from network patterns, enabling more accurate identification of truly critical events that warrant operator attention.
3Adaptability or versatility
If manually crafted rules are used for event prioritization, then rules can be tailored to network topology, but the rules must be updated when topology changes
Solution Approach 1:
The patent implements dynamic adaptability through ML models that automatically learn from changing network conditions. The system continuously processes new network event data to update its understanding of network topology and relationships, enabling it to adapt to changes without manual rule updates.
Solution Approach 2:
The system incorporates feedback mechanisms where network events and their resolutions are used to continuously improve the ML model's predictions. This feedback loop enables the system to automatically adjust to changing network topologies and patterns without requiring manual rule maintenance or updates.
4Reliability
If network operators address all critical events, then network reliability is maintained, but the overwhelming volume of events prevents timely response
Solution Approach 1:
The system extracts and prioritizes only the most time-sensitive critical events that pose immediate risk to network stability. By focusing operator attention on these extracted high-priority events rather than all critical events, the system maintains network reliability while enabling timely response to the most important issues.
Data Source
AI summary
Event management system and method. Events comprise physical and logical attributes. Tuples are created to identify a set of logical attributes. The tuples are arranged in hierarchized relations by creating binarized co-occurrence matrices, each co-occurrence matrix reflecting different time intervals and indicate occurrence of tuples in time windows of the time intervals. Tuple pairs are analyzed to determine probabilistic score related to co-occurrence, and tuple families are created from tuple pairs based on the probabilistic score. From tuple families, events are used to extract tuple instances including physical attributes, which are arranged as tuple-instance families using the corresponding tuple families as reference.