Figure 1 gives a usage example of SVDFeature to implement the model introduced in previous
section. We encode the album information as item features and item day biases as global features.
The number 0.6 associated with album id is an empirical parameter chosen by the user to control
the influence of album information in prediction. Assuming there are only two days’ records in the
data, the global feature index is defined as gid = 2×iid +day,where the number of item day biases
is twice the number of items. SVDFeature will learn a feature-based matrix factorization model with
the given training data and make predictions on supplied test feature files. We provide a manual to
give more details about the usage of SVDFeature