Combining content and ratings in hybrid recommendation engines
Instead of seeing rating-based recommenders as a successor to content-based recommenders, you should consider them as a different recommender after having acquired enough user-item interaction data to provide rating-only recommendations. In most practical cases, a recommendation engine will exist for both approaches – either as two distinct algorithms or a single hybrid model. In this section, we will look into training such a hybrid model.
To build a state-of-the-art recommender using the Matchbox recommender, open Azure Machine Learning designer and add the building blocks for the Matchbox recommender to the canvas, as shown in the following diagram. As we can see, the recommender can now take ratings and user and item features as input to create a hybrid recommendation model:
Figure 13.9 – The Matchbox recommender in Azure Machine Learning designer
In order to configure the Matchbox...