Accès libre

Academic Collaborator Recommendation Based on Attributed Network Embedding

Ouxia Du

Ouxia Du

School of Computer and Information Science, Southwest UniversityChina
Recherchez cet auteur sur
Sciendo|Google Scholar
Du, Ouxia
 et 
Ya Li

Ya Li

School of Computer and Information Science, Southwest UniversityChina
Recherchez cet auteur sur
Sciendo|Google Scholar
Li, Ya
  
03 févr. 2022
Academic Collaborator Recommendation Based on Attributed Network Embedding's Cover Image
À propos de cet article
Article précédent
Article suivant
Citez
Télécharger la couverture
Catégorie d'article: Research Paper
Publié en ligne: 03 févr. 2022
Pages: 37 - 56
Reçu: 10 nov. 2021
Accepté: 11 janv. 2022
DOI: https://doi.org/10.2478/jdis-2022-0005
Mots clés
© 2022 Ouxia Du et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1

The framework of our proposed ACR-ANE model.
The framework of our proposed ACR-ANE model.

Figure 2

Capture non-local neighbors.
Capture non-local neighbors.

Figure 3

Capture attr_sim neighbors.
Capture attr_sim neighbors.

Figure 4

Preservation of multi-type academic relationships.
Preservation of multi-type academic relationships.

Figure 5

Influence of Freq on Precision, Recall, and F1(Aminer).
Influence of Freq on Precision, Recall, and F1(Aminer).

Figure 6

Influence of Freq on Precision, Recall and F1(APS).
Influence of Freq on Precision, Recall and F1(APS).

Figure 7

Influence of scholar embedding dimension on Precision, Recall, and F1(Aminer).
Influence of scholar embedding dimension on Precision, Recall, and F1(Aminer).

Figure 8

Influence of scholar embedding dimension on Precision, Recall, and F1(APS).
Influence of scholar embedding dimension on Precision, Recall, and F1(APS).

Figure 9

Comparison between ACR-ANE and baselines in terms of Precision, Recall, and F1(Aminer).
Comparison between ACR-ANE and baselines in terms of Precision, Recall, and F1(Aminer).

Figure 10

Comparison between ACR-ANE and baselines in terms of Precision, Recall, and F1(APS).
Comparison between ACR-ANE and baselines in terms of Precision, Recall, and F1(APS).

Notations_

Notation Description
G The attributed academic collaboration network
V The set of all scholars
E The set of relationship between scholars
S The weight of edge
A Scholar attribute matrix
X Adjacency matrix of multi-relational networks
Y Final scholar embedding matrix
d Scholar embedding dimension
da Scholar attribute embedding dimension
xi, x^i {{{\boldsymbol{\hat x}}}_{\boldsymbol{i}}} The input data and reconstructed data
W(k), Ŵ(k) The k-th layer weight matrix
b(k), b^(k) {{{\boldsymbol{\hat b}}}^{{\boldsymbol{(k)}}}} The k-th layer biases

Statistics of two datasets_

Datasets # of Nodes # of Links
Aminer 7,436 11,568
APS 5,102 39,333
Langue:
Anglais
Périodicité:
4 fois par an
Sujets de la revue:
Informatique, Informatique, Gestion de projet, Bases de données et exploration de données
RSS Feed de la revue