Train a Machine Learning Model

Configure CSM to run a scheduled task that trains a machine learning model. This model analyzes an input field and generate a prediction.

Task Notes
1. Configure a machine learning model

Create a machine learning model using the machine learning Manager. This will determine the field that the model will analyze and the field that will be predicted.

See Configure a Machine Learning Model.

2. Create a query to provide example data

Use the Search Manager to configure a machine learning query that returns example content for the input field and the prediction field. To create a model that makes quality predictions that are both precise and accurate, the sample data returned by this query must provide a large amount of verified data.

See Searching.

3. Schedule the training

Create a Schedule Item that will use the machine learning search to pull the appropriate data. The Action Train Machine Learning uses that data to train the machine learning model. When the Schedule Item runs, it processes the data returned by the stored search using the algorithm to train the model. This training can be set as a recurring event. As new data is saved to CSM, it is used to retrain the model, and the quality of the predictions can improve.

See Scheduler.

4. Create an expression

Create a machine learning Prediction Expression to autopopulate the prediction and confidence score fields.

See Expressions.

5. Configure the fields

Configure a Business Object to include the input field and the prediction field specified in the machine learning model. The confidence score for a prediction can be displayed in a form as well.

See Define Auto-population Properties for a Field.