Hierarchical latent variable model-based news recommendation method

Abstract

The invention discloses a hierarchical latent variable model-based news recommendation method. The method comprises the steps of firstly extracting different types of news from major portal websites; then according to tags of the news, extracting news features; for registered users of a system, extracting explicit features and implicit features of the users to form user features; according to the user features and the news features, clustering the users and the news; adopting different latent variable models for the clustered news and users; predicting scores of the users to the news; selecting out multiple pieces of the news with the highest scores; and recommending personalized news to the users.

Description

technical field [0001] The invention relates to a news recommendation method based on a layered latent variable model. Background technique [0002] Recommender systems can be roughly divided into three categories: content-based recommendation, collaborative filtering recommendation, and hybrid recommendation systems. The content-based recommendation method originated from information retrieval and information filtering. It searches based on the similarity of content, resulting in recommended content that is too similar and lacks novelty; the recommendation method based on collaborative filtering is the earliest user mail filtering and document filtering, based on user The similarity of users or items in the rating matrix is ​​recommended, but in practice, the rating matrix of users and items is very sparse, resulting in inaccurate recommendations; while the hybrid recommendation system adds user feature information and item feature information to the linear model, so that c...

AI Technical Summary

Problems solved by technology

The content-based recommendation method originated from information retrieval and information filtering. It searches based on the similarity of the content, resulting in the recommended content being too similar and lacking in novelty; the recommendation method based on collaborative filtering is the earliest user email filtering and document filtering. The similarity of users or items in the scoring matrix is ​​recommended, but in practice, the user-item rating matrix is ​​very sparse, resulting in inaccurate recommendations; while the hybrid recommendation system adds user feature information and item feature information to the linear model, so that content-based The information of recommendation and collaborative filtering is fused together, which not only retains the characteristics of users and items in content-based filtering, but also adds information about the scoring matrix in collaborative filtering

[0003] In this mixed model, the user characteristics and item characteristics partly construct a linear model, but it is unreasonable to simply construct a linear model for the user group or item group, because forcibly using a linear model to fit nonlinear data will lead to a lot of problems. big error

Images