Making recommender system predictions with DataRobot

Creating suggestions from recommendation engines on DataRobot is straightforward. We use the drag and drop approach (as discussed in earlier chapters), as our prediction dataset is only small. With larger datasets (over 1 GB), as is more typical for recommender systems, using the DataRobot prediction API is advised. The API approach to creating models and making predictions is covered in depth in Chapter 12, DataRobot Python API.

Our prediction dataset for our example is 64 MB in size, and so the drag and drop approach is appropriate. For this prediction approach, we specify the columns we want to use from the original dataset. Ideally, we at least need an identifier for the item and user. As illustrated in Figure 10.7, we have chosen to include the ISBN, user_id, and title fields in our predictions. We drag and drop the prediction dataset into the specified region. As usual, this dataset is quickly evaluated, and we are presented...