Hierarchical attention time-series (hat) model for behavior prediction

Abstract

Aspects of the present disclosure provide techniques for behavior prediction. Embodiments include receiving activity data of a user, identifying user sessions comprising sets of time-stamped actions in the activity data, and segmenting the activity data into subsets corresponding to the user sessions. Embodiments include providing the subsets as inputs to a hierarchical attention time-series (HAT) model comprising: a first layer that determines attention scores for respective time-stamped actions in the subsets; and a second layer that determines attention scores for the subsets based on aggregations of the attention scores for the respective time-stamped actions. Embodiments include receiving, as outputs from the HAT model in response to the inputs: a prediction based on the subsets, the attention scores for the respective time-stamped actions, and the attention scores for the subsets; and explanatory information based on the attention scores for the respective time-stamped actions and the attention scores for the subsets.