Enhancing Anchor Link Prediction in Information Networks through Integrated Embedding Techniques

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F23%3A10253781" target="_blank" >RIV/61989100:27240/23:10253781 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.webofscience.com/wos/woscc/full-record/WOS:001063574700001" target="_blank" >https://www.webofscience.com/wos/woscc/full-record/WOS:001063574700001</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.ins.2023.119331" target="_blank" >10.1016/j.ins.2023.119331</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Enhancing Anchor Link Prediction in Information Networks through Integrated Embedding Techniques

Popis výsledku v původním jazyce

There are multiple types of information networks, including: social networks, citation networks, email communications networks, etc. are becoming popular in recent years. They have attracted a lot of researchers in multiple disciplines. Within this research domain, network alignment is considered as one of active topic due to its potential applications in real-world systems. In fact, with the popularity and diversity of information network types, most of users may participate in multiple information networks for many purposes. Thank to tremendous grow in the number of information networks recently, data analysis in information network has become a major challenge in multiple scientific disciplines. Anchor link prediction, also known as network alignment, which is aimed to match the users between different networks who have the same identification, is one of the emerging research directions in this domain. However, recent anchor link prediction baselines still lack the capability of sufficiently preserving global graph-structured features of individual information networks to obtain better prediction results. To overcome this challenge, we propose a novel model to align users between information networks using a seed set of known anchor links. To achieve better representations of individual networks in our proposed model, four embedding techniques are combined and learnt from the same latent space. We then apply the aggregation method to achieve a final network aligned embedding matrix which is later utilized to deal with the anchor link prediction problem. We run comprehensive experiments in real-life network alignment datasets to evaluate the effectiveness and compare them with the up-to-date baseline methods. (C) 2023

Název v anglickém jazyce

Enhancing Anchor Link Prediction in Information Networks through Integrated Embedding Techniques

Popis výsledku anglicky

There are multiple types of information networks, including: social networks, citation networks, email communications networks, etc. are becoming popular in recent years. They have attracted a lot of researchers in multiple disciplines. Within this research domain, network alignment is considered as one of active topic due to its potential applications in real-world systems. In fact, with the popularity and diversity of information network types, most of users may participate in multiple information networks for many purposes. Thank to tremendous grow in the number of information networks recently, data analysis in information network has become a major challenge in multiple scientific disciplines. Anchor link prediction, also known as network alignment, which is aimed to match the users between different networks who have the same identification, is one of the emerging research directions in this domain. However, recent anchor link prediction baselines still lack the capability of sufficiently preserving global graph-structured features of individual information networks to obtain better prediction results. To overcome this challenge, we propose a novel model to align users between information networks using a seed set of known anchor links. To achieve better representations of individual networks in our proposed model, four embedding techniques are combined and learnt from the same latent space. We then apply the aggregation method to achieve a final network aligned embedding matrix which is later utilized to deal with the anchor link prediction problem. We run comprehensive experiments in real-life network alignment datasets to evaluate the effectiveness and compare them with the up-to-date baseline methods. (C) 2023

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10200 - Computer and information sciences

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Information sciences

ISSN

0020-0255

e-ISSN

1872-6291

Svazek periodika

645

Číslo periodika v rámci svazku

říjen 2023

Stát vydavatele periodika

US - Spojené státy americké

Počet stran výsledku

14

Strana od-do

—

Kód UT WoS článku

001063574700001

EID výsledku v databázi Scopus

2-s2.0-85162883482