Visually-aware Interpretability Recommendation:Theory And Applications

A recommendation system is a computer program that identifies a particular object based on different user interests.The recommendation system has been an active research field for decades;the widespread adoption of the latent factor model makes the interpretability of recommendations a key issue in the research community and in practical applications.The lack of interpretability in the recommendation system can have a negative impact on the credibility of the recommendation system,thus affecting the effectiveness of the recommendation engine.In order to improve the accuracy and scalability of prediction,the invisible factor model based on matrix decomposition has been widely used in real-world systems.In this paper,based on the discussion and analysis of dataset features,existing models and experimental results,an implicit factor model based on visual interpretation and a dynamic user preference optimization recommendation model based on tense are proposed.This article has mainly completed the following work:.1.Introduce the related technology and application background of recommendation system,especially the implementation of single-class collaborative filtering algorithm based on matrix decomposition;2.Based on the analysis and comparison of the current collaborative filtering algorithm,the limitations of the algorithm,such as the sparsity feature of the data set and the cold start problem,are illustrated.3.Focus on the modeling methods considering the visual and non-visual factors of commodities,and introduce the time dynamics characteristics of commodities,and propose a recommendation model based on commodity hierarchy categories and fashion factors;4.The two optimization models proposed,namely the learning methods of R-TVPBR and T-sherlock model,are discussed,and the main C++ code implementation is given.5.Based on Amazon's five data sets,the R-TVPBR and T-sherlock optimization models are compared with other BPR-MF,VBPR,VBPR-C and Sherlock models to verify the effectiveness of the proposed algorithm.