Machine Learning
In IBM Process Mining, you can use machine learning to get predictions on the outcome of processes in terms of lead time and total cost. You can also get information about process steps that have not run yet. This additional information can help you to react in advance to potential undesired process behaviors.
Select Machine Learning on the left pane.
Note: Machine Learning is available on the Manage tab only if the following conditions are true:
- IBM Process Mining is installed on Red Hat OpenShift Container Platform.
- Machine learning is enabled in IBM Process Mining during installation. If machine learning is not enabled, contact your system administrator.
- The process is a flat process and not a multi-level process.
- You are the owner of the process.
Note: The Machine Learning page is visible only if you are using an owned process. This page is not available in snapshot or shared processes. However, the machine learning predictions in the Analytics page is available for both owned and shared processes.
You can train machine learning models to predict the outcome of the running processes in terms of Lead Time and Total Cost. A Training Required status is displayed to indicate that the model needs training. Click Train Model to start training the required KPI prediction model.
The following table explains the elements in Machine Learning section.
| Element | Description |
|---|---|
| Lead Time | Indicates the status of the model that estimates the total duration of running processes. |
| Total Cost | Indicates the status of the model that estimates the total cost of running processes. |
| Status | Indicates the status of model. The following statuses might be displayed for each KPI: - Training required: Indicates that machine learning requires training. For more information, go to Training Models. - In queue: Indicates that the model is in queue for training. - Training: Indicates that the model training is in progress. - Predicting: Indicates that the model is trained and is predicting the outcomes. - Training completed: Indicates that the model training is complete. - Failed: Indicates that the training has failed. You can hover over the status to see more information. |
| Train Model | Click here to train the machine learning model. |
Training models
Models require training in any of the following situations:
- Training was never done before.
- Data was modified or removed from the data source after the last training.
- End activities were changed (causing a different division of completed and running processes) after the last training.
- Mapping in the data source was changed after the last training.
- Costs or working times were changed after the last training.