Machine Learning functions
BMC Helix Edge leverages Machine Learning (ML) for early detection of issues. You can use this feature to reduce downtime, increase operational efficiency, perform predictive maintenance, and improve remediation. The ML algorithms analyze device metrics, attributes, and contextual data to identify patterns and unusual behavior. By spotting trends and anomalies, ML helps prevent outages and service interruptions and quickly identifies root causes of issues to reduce resolution time. It optimizes resource management by predicting. ML also enhances security by detecting suspicious activities and effectively responding to threats.
The following diagram displays the data flow in the BMC Helix Edge data stream system:
The following table describes the components and their functions in the BMC Helix Edge data stream:
Component name | Description |
|---|---|
Device service | This service generates a raw metric stream based on data collected from various devices. The raw metrics include multiple data types, such as temperature and RPM. |
Enrichment pipeline | This pipeline enriches the raw metric data with additional device metadata and configuration details. The enriched data is stored in a time-series metric data store and the BMC Helix environment. |
Pipeline clients | The pipeline clients apply rules and machine learning predictions to the enriched metric data. The data flows through the rule execution and ML prediction stages to generate the BMC Helix Edge event stream. |
Event clients | The event clients handle the integration with BMC Helix Operations Management and the remediation workflow defined in the workflow builder. |
Command executors | The command executors take actions according to the data processed. For example, if the processed data contains issues, the executors create tickets for the identified issues. |
BMC Helix Edge Dashboards | The BMC Helix Dashonards represent data visually for analysis and monitoring. |
Redis and other external systems cache and fetch additional data required in the pipeline, ensuring a comprehensive and informed data processing and action execution mechanism.
Supported algorithms types
The following list describes the support algorithm types with their purposes:
Algorithm type | Purpose |
|---|---|
Anomaly Detection | Identifies unusual patterns or situations that deviate from expected behavior based on historical data. This algorithm generates a score representing the anomaly level, typically ranging from Normal to Very Abnormal. |
Classification | Supports single-class classification. |
Regression | Predicts the value of a target metric based on other input metrics. |
Time Series Prediction | Uses historical data to predict future values of a time-dependent metric. |