Personalized Ranking Point of Interest Recommendation Based on Spatial-Temporal Distance Metric in LBSNs

Abstract

Nowadays, with the improvement of social network check-in and positioning technology, the positioning information is more accurate, and a large amount of network check-in data is generated. The recommendation research of interest points based on social networks is also increasing. Most of the points of interest refer to rely on geography, time, space, and textual information. In spatial-temporal, most studies consider the check-in rules from the geographical distance and time series. This paper introduces a geographic spatial-temporal distance measurement model to map temporal space information into a three-dimensional elliptical spherical coordinate system. The spatial-temporal distance is measured under the same reference standard. Helps alleviate the problems caused by cold start and data sparseness for location recommendation accuracy. Based on the Bayesian personalized ranking, this paper measures the temporal and spatial distance by using a Gaussian kernel function to weight the spatial-temporal distance, and proposes a personalized ranking recommendation algorithm based on the spatial-temporal distance metric. And it performs well on both datasets and is superior to the benchmark method.