Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

Abstract

Explainable recommender systems aim to generate explanations for users according to their predicted scores, the user’s history and their similarity to other users. Recently, researchers have proposed explainable recommender models using topic models and sentiment analysis methods providing explanations based on user’s reviews. However, such methods have neglected improvements in natural language processing, even if these methods are known to improve user satisfaction. In this paper, we propose a neural explainable collective nonnegative matrix factorization (NECoNMF) to predict ratings based on users’ feedback, for example, ratings and reviews. To do so, we use collective non-negative matrix factorization to predict user preferences according to different features and a natural language model to explain the prediction. Empirical experiments were conducted in two datasets, showing the model’s efficiency for predicting ratings and generating explanations. The results present that NECoNM F improves the accuracy for explainable recommendations in comparison with the state-of-art method in 18.3% for NDCG@5, 12.2% for HitRatio@5, 17.1% for NDCG@10, and 12.2% for HitRatio@10 in the Yelp dataset. A similar performance has been observed in the Amazon dataset 7.6% for NDCG@5, 1.3% for HitRatio@5, 7.9% for NDCG@10, and 3.9% for HitRatio@10.
Original languageUndefined/Unknown
Title of host publicationProceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.
EditorsMaría José Escalona, Francisco José Domínguez Mayo, Tim A. Majchrzak, Valérie Monfort
Number of pages11
PublisherSCITEPRESS Digital Library
Publication date2018
Pages35-45
ISBN (Print)978-989-758-324-7
DOIs
Publication statusPublished - 2018
Event14th International Conference on Web Information Systems and Technologies - Sevilla, Spain
Duration: 18 Sep 201820 Sep 2018

Conference

Conference14th International Conference on Web Information Systems and Technologies
CountrySpain
CitySevilla
Period18/09/201820/09/2018

Cite this

Costa, F., & Dolog, P. (2018). Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems. In M. J. Escalona, F. J. D. Mayo, T. A. Majchrzak, & V. Monfort (Eds.), Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018. (pp. 35-45). SCITEPRESS Digital Library. https://doi.org/10.5220/0006893700350045
Costa, Felipe ; Dolog, Peter. / Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems. Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.. editor / María José Escalona ; Francisco José Domínguez Mayo ; Tim A. Majchrzak ; Valérie Monfort. SCITEPRESS Digital Library, 2018. pp. 35-45
@inproceedings{c5348cd2fa184d39ad170c0ff6e5ae2e,
title = "Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems",
abstract = "Explainable recommender systems aim to generate explanations for users according to their predicted scores, the user’s history and their similarity to other users. Recently, researchers have proposed explainable recommender models using topic models and sentiment analysis methods providing explanations based on user’s reviews. However, such methods have neglected improvements in natural language processing, even if these methods are known to improve user satisfaction. In this paper, we propose a neural explainable collective nonnegative matrix factorization (NECoNMF) to predict ratings based on users’ feedback, for example, ratings and reviews. To do so, we use collective non-negative matrix factorization to predict user preferences according to different features and a natural language model to explain the prediction. Empirical experiments were conducted in two datasets, showing the model’s efficiency for predicting ratings and generating explanations. The results present that NECoNM F improves the accuracy for explainable recommendations in comparison with the state-of-art method in 18.3{\%} for NDCG@5, 12.2{\%} for HitRatio@5, 17.1{\%} for NDCG@10, and 12.2{\%} for HitRatio@10 in the Yelp dataset. A similar performance has been observed in the Amazon dataset 7.6{\%} for NDCG@5, 1.3{\%} for HitRatio@5, 7.9{\%} for NDCG@10, and 3.9{\%} for HitRatio@10.",
author = "Felipe Costa and Peter Dolog",
year = "2018",
doi = "10.5220/0006893700350045",
language = "Udefineret/Ukendt",
isbn = "978-989-758-324-7",
pages = "35--45",
editor = "Escalona, {Mar{\'i}a Jos{\'e}} and Mayo, {Francisco Jos{\'e} Dom{\'i}nguez} and Majchrzak, {Tim A.} and Val{\'e}rie Monfort",
booktitle = "Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.",
publisher = "SCITEPRESS Digital Library",

}

Costa, F & Dolog, P 2018, Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems. in MJ Escalona, FJD Mayo, TA Majchrzak & V Monfort (eds), Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.. SCITEPRESS Digital Library, pp. 35-45, 14th International Conference on Web Information Systems and Technologies , Sevilla, Spain, 18/09/2018. https://doi.org/10.5220/0006893700350045

Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems. / Costa, Felipe; Dolog, Peter.

Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.. ed. / María José Escalona; Francisco José Domínguez Mayo; Tim A. Majchrzak; Valérie Monfort. SCITEPRESS Digital Library, 2018. p. 35-45.

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems

AU - Costa, Felipe

AU - Dolog, Peter

PY - 2018

Y1 - 2018

N2 - Explainable recommender systems aim to generate explanations for users according to their predicted scores, the user’s history and their similarity to other users. Recently, researchers have proposed explainable recommender models using topic models and sentiment analysis methods providing explanations based on user’s reviews. However, such methods have neglected improvements in natural language processing, even if these methods are known to improve user satisfaction. In this paper, we propose a neural explainable collective nonnegative matrix factorization (NECoNMF) to predict ratings based on users’ feedback, for example, ratings and reviews. To do so, we use collective non-negative matrix factorization to predict user preferences according to different features and a natural language model to explain the prediction. Empirical experiments were conducted in two datasets, showing the model’s efficiency for predicting ratings and generating explanations. The results present that NECoNM F improves the accuracy for explainable recommendations in comparison with the state-of-art method in 18.3% for NDCG@5, 12.2% for HitRatio@5, 17.1% for NDCG@10, and 12.2% for HitRatio@10 in the Yelp dataset. A similar performance has been observed in the Amazon dataset 7.6% for NDCG@5, 1.3% for HitRatio@5, 7.9% for NDCG@10, and 3.9% for HitRatio@10.

AB - Explainable recommender systems aim to generate explanations for users according to their predicted scores, the user’s history and their similarity to other users. Recently, researchers have proposed explainable recommender models using topic models and sentiment analysis methods providing explanations based on user’s reviews. However, such methods have neglected improvements in natural language processing, even if these methods are known to improve user satisfaction. In this paper, we propose a neural explainable collective nonnegative matrix factorization (NECoNMF) to predict ratings based on users’ feedback, for example, ratings and reviews. To do so, we use collective non-negative matrix factorization to predict user preferences according to different features and a natural language model to explain the prediction. Empirical experiments were conducted in two datasets, showing the model’s efficiency for predicting ratings and generating explanations. The results present that NECoNM F improves the accuracy for explainable recommendations in comparison with the state-of-art method in 18.3% for NDCG@5, 12.2% for HitRatio@5, 17.1% for NDCG@10, and 12.2% for HitRatio@10 in the Yelp dataset. A similar performance has been observed in the Amazon dataset 7.6% for NDCG@5, 1.3% for HitRatio@5, 7.9% for NDCG@10, and 3.9% for HitRatio@10.

U2 - 10.5220/0006893700350045

DO - 10.5220/0006893700350045

M3 - Konferenceartikel i proceeding

SN - 978-989-758-324-7

SP - 35

EP - 45

BT - Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.

A2 - Escalona, María José

A2 - Mayo, Francisco José Domínguez

A2 - Majchrzak, Tim A.

A2 - Monfort, Valérie

PB - SCITEPRESS Digital Library

ER -

Costa F, Dolog P. Neural Explainable Collective Non-negative Matrix Factorization for Recommender Systems. In Escalona MJ, Mayo FJD, Majchrzak TA, Monfort V, editors, Proceedings of the 14th International Conference on Web Information Systems and Technologies, WEBIST 2018, Seville, Spain, September 18-20, 2018.. SCITEPRESS Digital Library. 2018. p. 35-45 https://doi.org/10.5220/0006893700350045