Train a Model on a Subset of Data

By default, topic models (like Theme Detection) will train on your entire data stream. However, users can also set filters when deploying a model to train the model on a specific subset of their data. The filter only applies to training – models will still analyze the entire dataset. 

To add a filter when deploying a model

1. In your workspace, navigate to the Settings tab, and select the data stream on which you want to deploy a model. 

2. In the center of the screen, select the topic model you wish to use. These can be found under "What are our customers saying? " In this example, we select Theme Detection. 

3. In the dialogue that opens, you can select your text & date dimensions as normal. To begin adding custom training filters, click "Switch to advanced setup." 

4. Scroll to the bottom of the updated dialogue, to "Custom Filters." Click "+ Custom Filter." 

5. In the dialogue that opens, you will see a sample of your data stream in a table view. Click "+ Add Filter" to select the field you wish to filter on. Once you are satisfied with your filter selections, click "Submit." Then, click "Start Analysis." 

6. Your model will now train on the subset of data defined by your selected filters. Click "Save" in the bottom right to begin processing.