2022 VOLUME 2 ISSUE 4
RESEARCH ARTICLE
Automatically identifying the motivations of Facebook mention to scholarly outputs based on Light GBM
AUTHOR
Houqiang Yu, Wei Zhang, Yue Wang, Yinghua Xie
ABSTRACT
[Purpose/Significance] The article investigated the automatic identification of the motivation of Facebook mention to scholarly outputs based on Light GBM algorithm, in order to achieve more in-depth usage of Facebook mention on a large scale. [Methodology/Procedure] Based on three types of contextual data, including mentioned scholarly outputs, Facebook users who post scholarly outputs, and text of Facebook posts to scholarly outputs, promising relevant features were extracted, and machine learning algorithms were used to automatically identify the motivations. [Results/Conclusions] (1) Features significantly correlated to the motivation of Facebook mention are identified in all three types of contextual data. In particular, relevant features are the altmetric attention score, the number of collaborative countries, the number of followers, the number of likes, the identities of Facebook users who post scholarly outputs and the number of comments on Facebook posts; (2) The prediction precision of the Light GBM classification model for motivation of Facebook mention was 0.31. In comparison, the classification precision without the text features of Facebook posts was 0.35, which was higher than the overall feature combination. The classification precision with only the post text features was 0.27. After combining the length and language of posts, the precision was improved to 0.30; (3) The classification precision of Facebook motivation has a positive correlation with users' activity. After combining all features, the classification precision of the first quartile users in terms of productivity reached 1, the classification precision of the second quartile was 0.36, and for the third quartile, the classification precision was 0.32. In conclusion, considering the high complexity of automatic classification of motivation of Facebook mentions, the study has achieved relatively high classification precision and could provide reference for future studies
KEYWORDS
Facebook mention; Facebook mention motivation; Automatic classification; Light GBM