Wireless Access Point Failure Prediction via Machine Learning Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network assurance systems face challenges in predicting wireless access point radio failures due to the complexity of network dynamics and the large number of parameters involved, leading to inefficiencies in proactive maintenance and user experience degradation.
Innovation Solution
A network assurance system that forms clusters of similarly behaving wireless access points, trains machine learning-based failure prediction models, and proactively triggers clients to roam to alternative access points based on predicted failures, while quarantining models when new software versions are introduced to ensure accurate predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If network assurance systems monitor and track numerous network metrics to assess network health, then the ability to detect and respond to network issues improves, but the system complexity and computational requirements increase significantly
Solution Approach 1:
The patent segments the network monitoring system into multiple specialized components: a data collection module that gathers metrics from multiple sources, a feature extraction module that identifies relevant patterns, a machine learning model training module that creates prediction models, and a failure prediction module that generates alerts. This segmentation allows each component to focus on specific tasks, improving overall reliability while managing complexity through modular architecture.
Solution Approach 2:
The system performs preliminary actions by continuously training machine learning models on historical network data and establishing baseline failure patterns before actual failures occur. The patent implements proactive model training and validation processes that prepare the system in advance, enabling it to predict failures rather than merely react to them, thus improving reliability while the automated preliminary processing helps manage complexity.
2Reliability
If the system uses machine learning models to predict access point failures, then proactive maintenance capability improves, but the difficulty of detecting and measuring failure patterns increases
Solution Approach 1:
The patent implements feedback mechanisms where the machine learning model continuously receives feedback from actual network performance data and failure outcomes. The system tracks prediction accuracy, compares predicted failures with actual failures, and uses this feedback to retrain and refine models. This feedback loop improves failure prediction accuracy while the automated feedback processing helps manage the complexity of pattern detection.
Solution Approach 2:
The system replaces manual analysis of complex failure patterns with machine learning algorithms. Instead of relying on human experts to detect subtle failure indicators in multidimensional network data, the patent employs trained models that automatically identify patterns, correlations, and anomalies. This substitution improves detection capability while the automation reduces the operational difficulty of measuring and analyzing failure patterns.
3Stability of the object's composition
If the system proactively triggers client roaming based on failure predictions, then network stability improves, but the loss of time for client reconnection and potential false alarms increases
Solution Approach 1:
The system applies preliminary anti-action by triggering client roaming before the access point actually fails. The machine learning model predicts failures in advance, allowing the system to proactively relocate clients to alternative access points before service disruption occurs. This prevents the harmful effect of failure while the advance timing of the action avoids the need for emergency reconnection, thus maintaining network stability without significant time loss.
4Measurement precision
If the system quarantines machine learning models when new software versions are introduced, then prediction accuracy is maintained, but the adaptability to new software versions decreases
Solution Approach 1:
The patent implements a dynamic model management system where quarantine status is not permanent but conditional and time-based. When new software versions are detected, models are temporarily quarantined and retrained on data from the new version. The system dynamically adjusts model status based on validation performance, allowing transition from quarantine to active deployment once accuracy is confirmed. This dynamic approach maintains prediction accuracy while enabling adaptation to new software versions.
Data Source
AI summary
In one embodiment, a network assurance system that monitors a network forms a cluster of similarly behaving wireless access points (APs). The cluster includes APs associated with different software versions. The network assurance system trains a machine learning-based failure prediction model for the cluster based on a set of features of the APs in the cluster. The network assurance system proactively triggers a client in the network to roam from a first AP to a second AP, based on the failure prediction model predicting a failure of the first AP. The network assurance system quarantines the failure prediction model when a new software version is associated with one or more of the APs.