Follow this procedure to integrate a data model created in H2O.
- Build a model in H2O.
- Build a Yellowfin report with data you wish to generate predictions for.
- Drag in a new column that is of the same data type as the prediction to be generated. For example, if your model returns a categorical prediction such as gender, you would want to use a text dimension.
- Within the column formatting drop-down, click Advanced Function.
- Select the Statistical option from the drop-down and select the H2O feature.
- Provide the full URL to your H2O instance (including the port number, as shown in the example below) and click Connect.
- If Yellowfin is able to connect to this URL, a drop-down menu will appear representing all of the models that are available for prediction. If your model is not currently supported, it will not appear in the list. Select the model you wish to generate predictions from and click Load.
A list of columns will appear representing the specified file’s required input. Match these to columns in your report.
A maximum limit of 25 input fields applies. The system will not load a model which contains more inputs than the limit allows.
- Click on Save.
- Note: If there was a problem in generating predictions, check that you have provided valid columns of the same data type required by H2O, or look at the yellowfin.log file for a complete error description.