Probability matrix decomposition cold start recommendation method fusing attributes and semantics

Abstract

The invention discloses a probability matrix decomposition cold start recommendation method fusing attributes and semantics. The method comprises the steps of firstly, extracting user attribute information, project attribute information, project text information and user scoring information from a database; and modeling the attribute information and the semantic information by utilizing linear regression to predict potential characteristics, and taking a predicted value as a prior probability of probability decomposition, thereby realizing fusion of the attribute information and the semantic information into probability decomposition of a scoring matrix. According to the method, attribute information and semantic information can be effectively fused into probability matrix decomposition, the ubiquitous problems of cold start and sparsity in a recommendation system are solved, and the method has higher accuracy and low algorithm complexity and is suitable for processing large-scale data.

Description

technical field [0001] The invention relates to the field of cold-start recommendation, in particular to a probabilistic matrix decomposition cold-start recommendation method integrating attributes and semantics. Background technique [0002] With the rapid development of technologies such as cloud computing, big data, and the Internet of Things, various services and user data on the Internet have exploded. These big data contain rich value and great potential, which has brought transformative development to human society. How to quickly and effectively obtain valuable information from complex data to make personalized recommendations for users is the research of recommendation system. key problem. Personalized recommendation system has become a hot spot in academia and industry and has produced many related research results. The recommendation system is based on user preferences, interests, etc., through recommendation algorithms to mine items of interest to users (such a...

AI Technical Summary

Problems solved by technology

Although the probability matrix decomposition only uses a single user-item rating matrix for mining, it can obtain the latent characteristics of users and items to a certain extent, and overcome the sparsity problem, but it lacks sufficient semantic interpretation and cannot solve the cold start problem.

[0004] At present, the academic community has studied the problems of matrix decomposition. Document 1 (Wang Yang, Zhong Yong, Li Zhendong, et al. A scoring prediction algorithm that integrates semantic similarity and matrix decomposition [J]. Computer Applications, 2017 (z1) .) A matrix decomposition algorithm that integrates semantic similarity is proposed, the semantic similarity between items is calculated through ontology and the missing values ​​in the scoring matrix are filled, and then the filled scoring matrix is ​​decomposed to achieve recommendation, but the artificial cost of constructing ontology objects is relatively high, poor scalability

Document 2 (Chen Pinghua, Zhu Yu. A recommendation algorithm that combines knowledge graph representation learning and matrix decomposition [J]. Computer Engineering and Design, 2018, 39(10): 145-150.) proposes a fusion of knowledge graph representation learning and matrix The decomposition recommendation algorithm uses the knowledge map to calculate the semantic similarity between entities, and integrates it into the matrix decomposition, which enhances the effect of matrix decomposition at the knowledge level, but cannot complete the recommendation for new users or new items

Document 3 (Zhang, Yufang.(2015).Collaborative Filtering Algorithm Based on Item Semantic and UserCharacteristics.Journal of Information and Computational Science.12.4059-4067.10.12733 / jics20106139.) proposed a collaborative filtering algorithm that combines item semantics and user characteristics. The cold start problem is alleviated by the improved linear fusion of user similarity and item similarity for predictive scoring, but this method only considers the local scoring data and ignores the global impact

Images