Multi-behavior perception graph comparison recommendation method

The invention discloses a multi-behavior perception graph comparison recommendation method. The method comprises the following steps: (1) representation learning on an original view; (2) representation learning on a noise-free view; (3) performing multi-behavior perception modeling, including potential behavior modeling, behavior representation learning of the step E and behavior comparison learning of the step M; and (4) multi-task learning. According to the method, various behaviors of the user are well modeled, so that the interaction relationship between the user and the article is more effectively captured, and the representation is further improved; and meanwhile, the noise problem in comparative learning is solved, so that the representation learned by the recommendation method is more effective. 本发明公开了一种多行为感知的图对比推荐方法，包括：(1)原始视图上的表示学习；(2)无噪视图上的表示学习；(3)多行为感知建模，包含潜在行为建模、E步的行为表示学习和M步的行为对比学习；(4)多任务学习。本发明很好地建模了用户的多种行为，从而更加有效地捕捉了用户/物品之间的交互关系，使得表示得到进一步提升；同时，解决了对比学习中的噪声问题，使得本发明推荐方法学习到的表示更加有效。