Detailed information on the specification for scoring models generated by a particular Algorithm can be found in the Algorithms section of the H2O documentation.
In general, models can be scored by choosing the appropriate algorithm from the Score drop down menu. Each of these scoring processes require that the testing data be in the same format as the training data.
Often the objective is not just to build and test a model, but to leverage that model’s predictive capability. This feature can be accessed by going to the drop down menu Score and selecting Predict. Data need not include a column for the predicted variable, but should correspond to the training data in all other dimensions. Prediction output is a single column vector that corresponds row for row with the data set submitted for prediction. This column is assigned a .hex key, which can be found in the Admin menu under jobs. Prediction output can also be downloaded to the users working directory as a .csv file by simply clicking the CSV link at the top of the Prediction Results page.