Improving POI Recommendation through Collaborative Filtering by incorporating the Geographical Factor into the Similarity Calculation

In recent years, the popularity of Location-Based Social Networks (LBSNs) has surged among tourists
seeking to share their travel experiences with their social circles. Although these platforms generate vast
amounts of data, effectively utilizing this information to provide personalized recommendations poses
significant challenges. Point-Of-Interest (POI) recommendation systems have emerged as a promising
solution, leveraging data from LBSNs to suggest tailored destinations for tourists. Collaborative Filtering
(CF) has gained recognition as a widely adopted memory-based technique. By analyzing user similarities,
CF often uses similarity metrics to predict the likelihood of tourists visiting specific POIs. This study
introduces a novel method, called IUPJS (Incorporation of User Proximity in Jaccard Similarity), which
extends the traditional Jaccard index by integrating geographic proximity into the similarity calculation.
Experimental evaluations on a Foursquare data set indicate that the proposed IUPJS significantly
enhances the effectiveness of the recommendation system. This improvement is particularly evident in key
evaluation metrics, including precision, recall, F1-score, mAP, and NDCG, exceeding the performance of
traditional methods commonly employed in the literature.