OCTOPUS: Across-social network user identification via multi-category spatio-temporal trajectories

Abstract The user and data scales of social networks are experiencing rapid expansion, leading to the emergence of diverse across-social network applications. Within this landscape, across-social network user identification has evolved into a popular and growing research area. Current approaches pri...

Full description

Saved in:
Bibliographic Details
Main Authors: Yating Qu, Ling Xing, Kaikai Deng, Honghai Wu, Yue Ling, Deshun Jia
Format: Article
Language:English
Published: Springer 2025-07-01
Series:Journal of King Saud University: Computer and Information Sciences
Subjects:
Online Access:https://doi.org/10.1007/s44443-025-00129-9
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract The user and data scales of social networks are experiencing rapid expansion, leading to the emergence of diverse across-social network applications. Within this landscape, across-social network user identification has evolved into a popular and growing research area. Current approaches primarily rely on various user data to construct models that portray user profiles, aiming to create solutions for user identification. However, the integration of multi-category user data is constrained by the data sparsity, fragmentation, and asymmetry characteristics, resulting in a limitation to the accuracy of user identification. To this end, we propose OCTOPUS, which consists of three modules: (i) a spatio-temporal extraction module adopts forward and backward LSTM networks to enhance the capacity of characterize spatio-temporal sequences, followed by using an attention mechanism to output complete user spatio-temporal features; (ii) a temporal perception module extracts temporal features from fine-grained and coarse-grained levels; (iii) a spatial location module obtains the spatial distribution of users to output user trajectory pair characteristics from global space. We perform extensive verification on two large public datasets (i.e., Brightkite and Gowalla), and the results show that OCTOPUS is superior to the state-of-the-art methods in terms of accuracy/precision/recall rate, and comprehensive evaluation index (F1).
ISSN:1319-1578
2213-1248