Open AccessPersonalised recommendation algorithm based on covariance
Author(s) -
Cai Biao,
Huang Yusheng
Publication year - 2020
Publication title -
the journal of engineering
Language(s) - English
Resource type - Journals
ISSN - 2051-3305
DOI - 10.1049/joe.2019.1231
Subject(s) - novelty , recommender system , computer science , collaborative filtering , covariance , limiting , algorithm , variety (cybernetics) , diversity (politics) , data mining , machine learning , information retrieval , artificial intelligence , statistics , mathematics , mechanical engineering , philosophy , theology , sociology , anthropology , engineering
Recommender systems are of great significance in predicting the potential interesting items based on the target user's historical selections. However, the recommendation list for a specific user has been found changing vastly when the system changes. So far, a variety of personalised recommendation algorithms have been proposed and most of them are based on similarities, such as collaborative filtering and mass diffusion. Most of the algorithms aimed at recommendation's accuracy. Here, the authors proposed a new Cov (covariance) recommendation method based on correlation coefficients, which can provide precision of recommendation. First, this method can express the positive and negative correlation among random samples without knowing the distribution of items. Then by limiting the popular items in recommendation list, an improved method is proposed, CovH, which can improve the diversity and novelty in recommendation compared with Cov. By applying the two methods on realistic recommendation datasets, such as Movie‐Lens, Netflix The authors found that comparing with baseline methods, these two algorithms can not only ensure decent accuracy but also have good performance on diversity and novelty indeed.