DFIAM: deep factorization integrated attention mechanism for smart TV recommendation

Zhou, Y; Shen, X; Zhang, S; Yu, D; Xu, G

DFIAM: deep factorization integrated attention mechanism for smart TV recommendation

Zhou, Y Shen, X Zhang, S Yu, D Xu, G

Permalink

Publisher:: Springer Science and Business Media LLC
Publication Type:: Journal Article
Citation:: World Wide Web, 2021, 24, (5), pp. 1465-1481
Issue Date:: 2021-09-01

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

The embargo period expires on 27 Jul 2022

Adobe PDF

Download Accepted versionAdobe PDF (1.67 MB)

View on publisher's site

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Zhou, Y
dc.contributor.author	Shen, X
dc.contributor.author	Zhang, S
dc.contributor.author	Yu, D
dc.contributor.author	Xu, G https://orcid.org/0000-0003-4493-6663
dc.date.accessioned	2022-01-19T00:59:08Z
dc.date.available	2022-01-19T00:59:08Z
dc.date.issued	2021-09-01
dc.identifier.citation	World Wide Web, 2021, 24, (5), pp. 1465-1481
dc.identifier.issn	1386-145X
dc.identifier.issn	1573-1413
dc.identifier.uri	http://hdl.handle.net/10453/153330
dc.description.abstract	Users are frequently overwhelmed by their uninterested programs due to the development of smart TV and the excessive number of programs. For addressing this issue, various recommendation methods have been introduced to TV fields. In TV content recommendation, auxiliary information, such as users’ personality traits and program features, greatly influences their program preferences. However, existing methods always fail to take auxiliary information into account. In this paper, aiming at personality program recommendation on smart TV platforms, we propose a novel Deep Factorization Integrated Attention Mechanism (DFIAM) model, which fully takes advantage of users’ personality traits, program and interaction features to construct users’ preference representations. DFIAM consists of two components, FNN component and DMF component. By suitably exploiting auxiliary information, FNN component devises a feature-interaction layer to capture the low- and higher-order feature interactions, while DMF component has a field-interaction layer to acquire higher-order field interactions. The embedding layer is divided into two layers , including feature embedding layer and field embedding layer. The two components share the feature embedding layer to profile latent representations of user and program features to reduce learning parameters and computational complexity. And the field embedding layer calculated by feature embedding layer is the input of DMF component. Besides, hierarchical attention networks are applied to self-adapt the influence of each feature and effectively capture more important feature interactions. To evaluate the performance of the DFIAM model, extensive experiments are conducted on two real-world datasets from different scenarios. The results of our proposed model have outperformed the mainstream neural network-based recommendation models in terms of RMSE, MAE and R-square.
dc.language	en
dc.publisher	Springer Science and Business Media LLC
dc.relation.ispartof	World Wide Web
dc.relation.isbasedon	10.1007/s11280-021-00924-0
dc.rights	This is a post-peer-review, pre-copyedit version of an article published in World Wide Web. The final authenticated version is available online at: https://doi.org/10.1007/s11280-021-00924-0
dc.rights	info:eu-repo/semantics/embargoedAccess
dc.subject	0804 Data Format, 0805 Distributed Computing, 0806 Information Systems
dc.subject.classification	Information Systems
dc.title	DFIAM: deep factorization integrated attention mechanism for smart TV recommendation
dc.type	Journal Article
utslib.citation.volume	24
utslib.for	0804 Data Format
utslib.for	0805 Distributed Computing
utslib.for	0806 Information Systems
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	/University of Technology Sydney/Strength - AAI - Advanced Analytics Institute Research Centre
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology/School of Computer Science
utslib.copyright.status	open_access	*
utslib.copyright.embargo	2022-07-27T00:00:00+1000Z
dc.date.updated	2022-01-19T00:59:07Z
pubs.issue	5
pubs.publication-status	Published
pubs.volume	24
utslib.citation.issue	5

Abstract:

Users are frequently overwhelmed by their uninterested programs due to the development of smart TV and the excessive number of programs. For addressing this issue, various recommendation methods have been introduced to TV fields. In TV content recommendation, auxiliary information, such as users’ personality traits and program features, greatly influences their program preferences. However, existing methods always fail to take auxiliary information into account. In this paper, aiming at personality program recommendation on smart TV platforms, we propose a novel Deep Factorization Integrated Attention Mechanism (DFIAM) model, which fully takes advantage of users’ personality traits, program and interaction features to construct users’ preference representations. DFIAM consists of two components, FNN component and DMF component. By suitably exploiting auxiliary information, FNN component devises a feature-interaction layer to capture the low- and higher-order feature interactions, while DMF component has a field-interaction layer to acquire higher-order field interactions. The embedding layer is divided into two layers , including feature embedding layer and field embedding layer. The two components share the feature embedding layer to profile latent representations of user and program features to reduce learning parameters and computational complexity. And the field embedding layer calculated by feature embedding layer is the input of DMF component. Besides, hierarchical attention networks are applied to self-adapt the influence of each feature and effectively capture more important feature interactions. To evaluate the performance of the DFIAM model, extensive experiments are conducted on two real-world datasets from different scenarios. The results of our proposed model have outperformed the mainstream neural network-based recommendation models in terms of RMSE, MAE and R-square.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/153330