SRGCN: Social Relationship Graph Convolutional Network-Based Social Network User Geolocation Prediction

Abstract

In location-based applications such as online shopping, recommendation system and disaster monitoring, the prediction of user geolocation plays an important role. Existing methods mainly utilize textual content and social network for location prediction, which fail to fully mine and fuse the two types of information on the one hand, and on the other hand, it is difficult to predict the location of isolated users in social network. Therefore, we propose a hybrid learning model Social Relationship Graph Convolutional Neural Network (SRGCN) to jointly model text features and network structure, in addition to constructing a social relationship graph for isolated users based on text topic similarity, and further improving the accuracy of location prediction through social relationships. The experiments on a real-world benchmark dataset validate the effectiveness of the method, showing that SRGCN outperforms existing methods in terms of average distance error, median distance error, and prediction accuracy.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 7 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.