Multi-relationship perception temporal interaction network prediction method

Abstract

The invention discloses a multi-relationship perception temporal interaction network prediction method. The method comprises the following steps: (1) taking interaction in a temporal interaction network as a sample; (2) sequentially processing each interaction according to the interaction occurrence time, mining nodes having a historical interaction relationship, a common interaction relationshipand an interaction sequence similarity relationship with interaction nodes on the basis of historical interaction information, and constructing a local relationship graph before current interaction for the interaction nodes; (3) predicting the representation of the article before current interaction according to the representation of the user after last interaction and the neighbor-based representation of the user obtained through hierarchical multi-relation perception aggregation; (4) updating the representation of the interaction node according to the representation of the interaction node after the last interaction, the time interval between the last interaction and the current interaction and the representation based on neighbors; and (5) after the temporal interaction network prediction model is trained, predicting articles which may be interacted by the user by using the temporal interaction network prediction model after parameter optimization.

Description

technical field [0001] The invention relates to the field of temporal interaction network prediction, in particular to a multi-relationship-aware temporal interaction network prediction method. Background technique [0002] In many areas of real life, such as e-commerce (customers buy goods), education platforms (students participate in MOOC courses) and social networking platforms (users post in the community), users will interact with different items at different times, and users The interactions between and items form a temporal interaction network. Compared with static interaction networks, temporal interaction networks add attention to the interaction time. Temporal interaction network prediction refers to predicting which item a user will interact with before the interaction occurs, which is of great significance for tasks such as product recommendation, course recommendation, and community recommendation. [0003] Existing prediction methods based on temporal intera...

AI Technical Summary

Problems solved by technology

Although the prediction method that does not consider neighbor information models the attribute changes of interactive nodes, it ignores the influence of neighbor information

Existing prediction methods that consider neighbor information consider only nodes with historical interaction relationships as neighbor nodes, ignoring other relationship types in historical interaction information (co-interaction relationship, interaction sequence similarity relationship, etc.)

Images