Managing trained Machine Learning models

Fine-tune and manage trained Machine Learning models to achieve better understanding of the models, make necessary adjustments, or remove obsolete models. 

To access trained Machine Learning models

Accessing trained machine-learning models is crucial for various tasks in the machine-learning lifecycle, including testing, validation, deployment, and inference. 

• On the BMC Helix Edge page, select Machine Learning.
  The BMC Helix Edge page displays the list of trained machine-learning models.

  Field name	Description
  Name	Displays the name assigned to the trained Machine Learning model.
  Device profile	Displays the associated device profile for the trained Machine Learning model.
  Description	Displays a brief overview of the trained Machine Learning model for users to understand the purpose or functionality of the model at a glance.
  Model type	Displays a specific type or category of the trained Machine Learning model to help users understand the intended use and capabilities.
  Actions	Provides options to perform various operations related to the trained Machine Learning model. Depending on the available functionalities, these actions include the following options: Edit: Use this option to modify the trained Machine Learning model settings Duplicate: Use this option to create duplicate trained models. You can create duplicate models for testing different configurations or deployment scenarios without affecting the performance of the original trained model.

To edit a trained Machine Learning model

1. On the BMC Helix Edge page, select Machine Learning.
2. In the trained Machine Learning model list, locate a model to edit and select Actions > Edit.
   BMC Helix Edge displays the following panel:
   

1. On the Edit model panel, update the following information:

   Field name	Description
   Name	Displays the name of the trained Machine Learning model.
   (Optional) Description	Offers a summary of what the trained Machine Learning model seeks to achieve. For example, type This model aims to monitor and analyze the correlation between temperature and humidity in HVAC systems.
   Algorithm type	Displays Anomaly as the algorithm. This version supports only Anomaly.
   Algorithm	Displays HedgeAnomaly as the algorithm for this trained Machine Learning model from the list. This version supports only HedgeAnomaly.
   Model information	Displays additional metadata or parameters that pertain to the trained Machine Learning model. For example, Helix Edge AutoEncoder is based on an Anomaly detection Algorithm.
   Device profile	Associate the model with a specific device or set of devices. Select the appropriate device profile from the available options to make sure the trained Machine Learning model is compatible with the target hardware—for example, H-VAC (provides humidity and temperature data).
   Model fields	Select the features or variables that the model uses for training. Add or remove fields that serve as input features or target variables for the model. For instance, under Device Attributes, select deviceName, and under Metrics, select Temperature and Humidity.
   Data options	Set the collection interval and add filters to refine the data used for training. Collection interval and filters: Specify the interval for data collection and apply any necessary filters. For example, to train the collection time in the Time period field, enter 60 days; in the Sample Every field, enter 60 seconds. Merge conditions: Use this option to combine attributes from different devices.  Attribute field: Choose the attribute to focus on. For example, select Temperature.  Operator: Select the operation to apply. The available options are Contains, Equals, and Excludes. For example, select Contains. Value: Select a value from the list. For example, select Normal.
   Preview data	Review the historical data selected for training the model.  Note: This field is visible only after you save the changes.

2. Click Save. 

To duplicate a trained Machine Learning model

By duplicating a trained Machine Learning model, you can version, test, and scale the model. Additionally, create new versions of existing models, enabling tracking model iterations and improvements. Lastly, use duplicate models for testing different configurations or deployment scenarios without affecting the performance of the original trained model.

1. On the BMC Helix Edge page, select Machine learning.
2. In the trained Machine Learning model list, locate a model to duplicate and select Actions > Duplicate.
   BMC Helix Edge displays the following panel:
   
   
3. In the Duplicate panel, perform the following steps:
   1. In the Name field, rename the model name. 
   2. (Optional) In the Description field, describe the model.
      For example, to test this model with different configurations, describe the model as Testing this trained model with different configurations.
   3. Click Duplicate.