TN-MR: topic-aware neural network-based mobile application recommendation

• Date: 06 February 2024
• Pages: 159-175
• DOI: https://doi.org/10.1108/IJWIS-10-2023-0205
• Published date: 06 February 2024
• Subject Matter: Information & knowledge management,Information & communications technology,Information systems,Library & information science,Information behaviour & retrieval,Metadata,Internet
• Author: Junyi Chen,Buqing Cao,Zhenlian Peng,Ziming Xie,Shanpeng Liu,Qian Peng

TN-MR: topic-aware neural

network-based mobile

application recommendation

Junyi Chen,Buqing Cao,Zhenlian Peng,Ziming Xie,

Shanpeng Liu and Qian Peng

School of Computer Science and Engineering,

Hunan University of Science and Technology, Xiangtan, China

Abstract

Purpose –With the increasing number of mobile applications, eff‌iciently recommending mobile

applications to users has become a challenging problem. Although existing mobile application

recommendation approaches based on user attributes and behaviors have achieved notable

effectiveness, they overlook the diffusion patterns and interdependencies of topic-specif‌icmobile

applications among user groups. mobile applications among user groups. This paper aims to capture the

diffusion patterns and interdependencies of mobile applications among user groups. To achieve this, a

topic-aware neural network-based mobile application recommendationmethod, referred to as TN-MR, is

proposed.

Design/methodology/approach –In this method, f‌irst, the user representations are enhanced by

introducing a topic-aware attention layer, which captures both the topic context and the diffusion history

context. Second, it exploits a time-decay mechanism to simulate changes in user interest. Multitopic user

representations are aggregated by the time decay module to output the user representations of cascading

representations under multipletopics. Finally, user scores that are likely to download the mobile application

are predictedand ranked.

Findings –Experimentalcomparisons and analyses were conducted on the actual360App data set, and the

results demonstrate that the effectiveness of mobile application recommendations can be signif‌icantly

improvedby using TN-MR.

Originality/value –In this paper, the authors propose a mobile application recommendation method

based on topic-aware attention networks. By capturing the diffusion patterns and dependencies of mobile

applications, it effectively assistsusers in selecting their applications of interest from thousands of options,

signif‌icantlyimproving the accuracy of mobile applicationrecommendations.

Keywords Mobile application recommendation, Topic-aware attention networks,

Time decay mechanism, Cascading representations, Attention and time decay mechanism

Paper type Research paper

1. Introduction

With the proliferation of mobiledevices, the development of mobile applications has become

increasingly popular. Mobile applications (apps) not only make people’s lives more

convenient but also meet their diverse requirements. However, in recent years, the number

of mobile applications has increased, and the market has f‌looded with mobile applications.

The work of this paper is supported by the National Natural Science Foundation of China withGrant

no. 62376062and 62177014, the National KeyR&D Program of China with Grantno. 2018YFB1402800,

Hunan ProvincialNatural Science Foundationof China with Grant no. 2022JJ30020and the Science and

TechnologyInnovation Program of Hunan Provincewith Grant No. 2023sk2081.

Mobile

application

recommendation

159

Received28 October 2023

Revised30 November 2023

Accepted4 December 2023

InternationalJournal of Web

InformationSystems

Vol.20 No. 2, 2024

pp. 159-175

© Emerald Publishing Limited

1744-0084

DOI 10.1108/IJWIS-10-2023-0205

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1744-0084.htm

This problem causes users to be overwhelmed by the vast number of mobile apps,

preventing them from swiftly and accurately locating the apps they need and increasing

their burden. Therefore, f‌iltering out the apps that users may be interested in from the

multitude of mobile apps and making personalized recommendations becomes a big

challenge. The mobile application recommendation scenario can be described as follows

[Figure 1(a)]: based on the user’s historicalapplication usage, download sequences, temporal

and other information, a model is trained to assess the similarity between the application

and users. Then, by identifyingthe next user who is most likely to meet these conditions, the

models recommend apps to that user. Currently, there are numerous approaches for

recommending mobile applications. Some studies use the collaborative f‌iltering (Dou et al.,

2016) algorithm to complete the prediction of the next application or user by capturing the

user’s historical behavioral characteristics and have achieved good results. Other studies

have attained mobile application recommendations using techniques such as matrix

decomposition (Koren et al., 2009) and logistic regression (Wang et al.,2016). With the

advancement of deep learning technology, neural network-based methods for

recommending mobile applicationshave become increasingly popular. Among these, NGCF

(Wang et al.,2019) is a representativemodel that combines traditional methods with neural

network technology. NGCF integrates deep learning technology and collaborative f‌iltering

Figure 1.

An illustrationof

traditionalmodeling

and topic-aware

modeling for mobile

application

recommendation

IJWIS

20,2

160