Based on long-term and short-term preferences for POI Recommendation

Point of interest (POI) recommendation is one of the important applications of social networks based on geographical location. Existing research methods tend to capture user preferences from recent check-in sequences, but they overlook the diverse and changing geographic relationships that can lead to changes in user interest preferences over time, resulting in a decrease in recommendation accuracy. To address these issues, we propose an integrated point of interest recommendation model that incorporates users' long-term and short-term preferences(ULSP). We utilize graph convolutional networks to learn the POI interaction graph and model users' long-term preferences to obtain vector representations of their historical activities. We calculate the time intervals between user check-ins and the spatial distances between check-in locations to capture their short-term preferences. Additionally, we employ self-attention mechanisms to assign different weights to user check-in data, thereby combining long-term and short-term interests to evaluate the probability of users visiting certain POIs in the future, thus achieving future POI recommendations. The experimental results show that the proposed method has improved on several indexes, and can effectively improve the accuracy of interest point recommendation.