Personalized region of interest recommendation through adaptive fusion of multi-dimensional user preferences
Abstract Region of interest (ROI) recommendation is essential for delivering personalized suggestions and optimizing resource allocation. This process involves analyzing users’ historical check-in data within location networks, which helps capture spatial activity preferences and predict regional mo...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Journal of Big Data |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40537-025-01224-4 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Region of interest (ROI) recommendation is essential for delivering personalized suggestions and optimizing resource allocation. This process involves analyzing users’ historical check-in data within location networks, which helps capture spatial activity preferences and predict regional movement patterns. To address challenges such as the limited diversity in behavioral distances between user regions, the lack of spatio-temporal correlation in non-contiguous regions, and an inadequate understanding of the variety of user preferences, we examine these preferences from three perspectives: spatio-temporal, social, and category. We then propose a multi-dimensional adaptive fusion method for personalized ROI recommendation. Firstly, the spatial intervals between potential ROIs are calculated by incorporating the shortest paths from geographic maps, directionality preferences, and regional activity patterns. These spatial intervals are then combined with time intervals to derive spatio-temporal preferences, utilizing a two-layer attention mechanism. Next, the social preferences are determined by assessing the influence of social connections and the impact of social networks on the likelihood of a user visiting a region, using a convolutional neural network. In addition, category features are extracted from the users’ historical check-in trajectories, and category preferences are calculated by evaluating the semantic similarity between the Point of Interest (POI) categories and user category features within a region using a multi-layer perceptron. Finally, an adaptive weighting model is introduced to integrate the spatio-temporal, social, and category preferences, assigning individual preference weights to different users to facilitate personalized ROI recommendation. We evaluated the proposed method using Foursquare and Sina Weibo datasets, in conjunction with the state-of-the-art baseline models. The results demonstrate that the proposed approach significantly enhances the critical performance metrics, including Recall, F1-score, and normalized discounted cumulative gain (NDCG). |
|---|---|
| ISSN: | 2196-1115 |