Hybrid recommendation method based on variational autoencoder

• Main Authors: LIN JIANSHI, ZHANG YIN
• Format: Patent
• Language: Chinese; English
• Published: 12.10.2018
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses a hybrid recommendation method based on a variational autoencoder. According to the method, a model is built for a score characteristic and a content characteristic of a user and an article by using a variational autoencoder, sparse characteristics are encoded through a factorization machine, and characteristic high-order combination is performed automatically; in addition,multi-view data characteristics of the user and the article are fused in the architecture of the variational autoencoder, so as to solve a problem of cold starting; variational inference analysis ofthe hidden vector code of the user and the article explains generation of the hidden vector code by the autoencoder; and preference values for candidate article collections of the user can be acquiredby inputting corresponding characteristics of the user and the article, and the candidate article collections are ranked according to preference values so as to obtain a recommendation result. Compared with traditional recomme