Time Series Event Prediction Using Weighted ML Ensembles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for predicting system events, such as outages, fail to identify complex and hidden intercomponent relationships between infrastructure components, leading to inefficient use of computing resources and inability to forecast critical events, resulting in system failures.
Innovation Solution
A method utilizing machine learning models to process historic time series data through block bootstrapping and hierarchical time series analysis, applying weights to predictions from multiple models to generate a final prediction, which enables the recognition of complex intercomponent relationships and critical events, allowing for preventative actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current prediction techniques are used, then computing resources are consumed, but system outages cannot be predicted and critical events are missed
Solution Approach 1:
The patent segments the time series data into multiple blocks and applies different machine learning models to different blocks. This segmentation allows the system to process data more efficiently while maintaining high prediction accuracy for critical events like system outages.
Solution Approach 2:
The patent dynamically assigns weights to different machine learning model predictions based on their performance. This dynamic weighting approach optimizes the use of computing resources by adaptively selecting the most reliable predictions while maintaining high overall prediction accuracy.
2Reliability
If multiple machine learning models are used with weighted aggregation, then prediction accuracy improves, but system complexity increases
Solution Approach 1:
The patent implements feedback mechanisms where prediction outcomes are used to dynamically adjust weights assigned to different machine learning models. This feedback loop improves prediction reliability while managing system complexity through automated adaptation rather than manual configuration.
Solution Approach 2:
The patent changes parameters (weights) of the machine learning models based on their performance. By dynamically adjusting these parameters, the system achieves higher prediction reliability without permanently increasing complexity, as the parameter changes are adaptive rather than structural.
Data Source
AI summary
A device may receive historic temporal data identifying events associated with a system, and may perform block bootstrapping of the hierarchical time series data, based on a hyperparameters, to generate blocks of data points of the historic time series data. The device may process the blocks of data points, with a plurality of different machine learning models, to calculate predictions, and may apply weights to the predictions to generate weighted predictions. The device may aggregate the weighted predictions to generate aggregated predictions, and may apply final weights to the aggregated predictions to generate weighted aggregated predictions. The device may aggregate the weighted aggregated predictions to generate a final prediction, and may perform one or more actions based on the final prediction.