Build Classification Models With Text

tell me more icon Tell me more

The text column is prepared and can be used to build models.

  1. In the Component Palette, expand the Models section. Click the Classification icon; move the mouse to the workflow and click again.

  2. The node is created. Rename the node TextClassification by selecting the name and typing the new name.

  3. Connect PrepareComments to TextClassification.

  4. The Edit Classification Build Node dialog opens. Select AFFINITY_CARD as the target and CUST_ID as the Case ID.

  5. Right-click TextClassification and select Run.

  6. Monitor the progress of the build in the Workflow Jobs.
    If the Workflow Jobs window is not open, open it using View > Data Miner > Workflow Jobs in the menu bar. Make sure that dmConnection is selected.

  7. After the build completes, you can examine the models that were built. You can see how the text features appear in the model build by viewing the models. For example, right-click TextClassification, select View Models and select the GLM model. The Coefficients tab shows coefficients for the tokens in COMMENTS_TOK
    tell me more icon Tell me more about what I see in the GUI.

  8. Right-click TextClassification, the classification node. Select Compare Test Results. Click Lift.

    Decision Tree (DT) models do not support Text; the COMMENTS column is ignored for the DT model.

    The other models support text; lift for these models is better than lift for text. Taking user comments into account results in better models.
    tell me more icon Tell me more about what I see in the GUI.

You can select a best model, if you wish. (Selecting a best model is described in Getting Started with Data Miner.) Alternatively, you can apply all four models.

tell me more icon Tell me more about what I see in the GUI.

Now apply the models to new data.

The cue cards icon next step is to prepare new data and apply the models to it.