Anomaly Detection for Application Infrastructure Using Machine Learning Models

The system segments the anomaly detection task into multiple specialized machine learning models, each trained on specific performance metrics (CPU utilization, memory utilization, response time, volume). This segmentation allows each model to focus on particular aspects of component behavior, improving detection accuracy while organizing complexity into manageable modular units that can be independently trained and deployed.

The patent introduces an intermediary layer between raw performance metrics and anomaly detection results. Machine learning models serve as intermediaries that process and interpret complex performance data, transforming raw metrics into meaningful anomaly detections. This intermediary layer handles the complexity internally while presenting simplified detection results to users.

The system performs preliminary action by training multiple machine learning models in advance on historical performance data. These pre-trained models are then deployed to make real-time anomaly detections without requiring extensive computational resources during operation. The heavy computational work is done beforehand, allowing reliable multi-model detection with lower runtime resource consumption.

The patent employs partial action by using multiple specialized models rather than one comprehensive model. Each model processes specific metrics partially, and their results are combined. This approach achieves reliable detection through the collective partial contributions of multiple models while optimizing resource usage compared to a single exhaustive model.