Analytic Engine for Converged Infrastructure Anomaly Detection

The patent transforms the evaluation approach by changing from traditional statistical parameters to machine learning-based parameters. The analytic engine uses supervised machine learning to predict expected performance values based on multiple input parameters (CPU usage, memory usage, response time, number of issues, customer service expense) and compares these predictions with actual values to identify true anomalies, thereby improving measurement precision while accounting for CI system complexity

The patent introduces an intermediary analytic engine that acts as a mediator between raw performance data and evaluation results. This engine processes performance data through machine learning models, incorporating normalization coefficients and statistically derived thresholds to filter out false alarms and produce accurate anomaly detection, thus resolving the contradiction between simplicity and accuracy

The patent applies preliminary action by pre-training the machine learning model with historical performance data to establish normalization coefficients and expected performance baselines. This pre-processing allows the system to make accurate predictions with reduced computational overhead during actual evaluation, as the model has already learned the relationships between input parameters and performance outcomes during the training phase

The analytic engine performs self-service by automatically adjusting normalization coefficients and updating its predictions based on incoming performance data. The system uses statistically derived thresholds that are automatically computed and applied, reducing the need for manual intervention and optimizing resource usage through adaptive learning rather than continuous heavy computation