tracking query rewrite activity for queries that are rewritten to access a materialized view;

storing object activity tracking data reflecting the query rewrite activity;

generating, based on the object activity tracking data, a database activity training set that contains a plurality of database activity examples, wherein each database activity example of the plurality of database activity examples contains a corresponding binary label that characterizes a database activity;

using the plurality of database activity examples to learn a classification model for the materialized view; and

using the learned classification model to predict whether query rewrite activity for queries rewritten to access the materialized view will occur during a future time interval;

wherein the method is performed by one or more computing devices.