Patent attributes
In some implementations, a method may involve determining first feature vectors for a plurality of data items accessed by a user of one or more client devices, the first feature vectors representing first contextual data about the one or more client devices at times that respective data items of the plurality of data items were accessed, the plurality of data items including a first data item. A predictive model, configured to classify input feature vectors into context types, may be used to determine that the first feature vector for the first data item is classified as a first context type. A second feature vector representing second contextual data about a first client device operated by the user may be determined and the predictive model may be used to determined that the second feature vector is classified as the first context type. Based at least in part on the first and second feature vectors being classified as the first context type and the first and second data items being of a first data item type, the first client device may be caused to present the second data item.