Open AccessRevisiting Local Walking Based on Social Network Trust (LWSNT): Friends Recommendation Algorithm in Facebook Social Networks
Author(s) -
Wahidya Nurkarim,
Arie Wahyu Wijayanto
Publication year - 2022
Publication title -
proceedings of international conference on data science and official statistics
Language(s) - English
Resource type - Journals
ISSN - 2809-9842
DOI - 10.34123/icdsos.v2021i1.124
Subject(s) - jaccard index , correlation , social network (sociolinguistics) , computer science , the internet , face (sociological concept) , order (exchange) , graph , social media , artificial intelligence , internet privacy , world wide web , theoretical computer science , mathematics , sociology , pattern recognition (psychology) , business , social science , geometry , finance
In the last decades, the internet penetration rate and online social network users have grown very fast. Online social network, such as Facebook, is a platform where one can find friends without having to meet face to face. A social network is represented by a large graph because it involves many participants. Hence, it is hard to find potential friends who have the same thoughts and interests. The Local Walking Based on Social Network Trust (LWSNT) algorithm is one of the popular algorithms for social friend recommendation. This study re-examines whether the correlation between attributes gives un-match ranks in different cases (cases with and without correlation). We assess the performance of LWSNT in Facebook networks under the supervised manner by comparing its F-score against similar methods. By using Kendall’s tau correlation, the results show that the correlation of attributes has no significant effect on the order of friend recommendations. In addition, the LWSNT performance is quite inferior against the Common Neighbors algorithm and Jaccard index.