Integrating spatial and temporal contexts into a factorization model for POI recommendation

ABSTRACT

Matrix factorization is one of the most popular methods in recommendation systems. However, it faces two challenges related to the check-in data in point of interest (POI) recommendation: data scarcity and implicit feedback. To solve these problems, we propose a Feature-Space Separated Factorization Model (FSS-FM) in this paper. The model represents the POI feature spaces as separate slices, each of which represents a type of feature. Thus, spatial and temporal information and other contexts can be easily added to compensate for scarce data. Moreover, two commonly used objective functions for the factorization model, the weighted least squares and pairwise ranking functions, are combined to construct a hybrid optimization function. Extensive experiments are conducted on two real-life data sets: Gowalla and Foursquare, and the results are compared with those of baseline methods to evaluate the model. The results suggest that the FSS-FM performs better than state-of-the-art methods in terms of precision and recall on both data sets. The model with separate feature spaces can improve the performance of recommendation. The inclusion of spatial and temporal contexts further leverages the performance, and the spatial context is more influential than the temporal context. In addition, the capacity of hybrid optimization in improving POI recommendation is demonstrated.

KEYWORDS:

• Check-in
• matrix factorization
• feature space separation
• POI recommendation

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The two data sets can be downloaded at http://www.ntu.edu.sg/home/gaocong/data/poidata.zip.

Additional information

Funding

This work was supported by the National Key Research and Development Program [grant number 2017YFB0503604], the NSFC Innovation Research Group Project [grant number 41421001]; the NSFC General Program [grant number 41371380, 41771477]; the Innovation Project of LREIS [O88RA20BYA] and the Key Programs of the Chinese Academy of Sciences [QYZDY-SSW-DQC007].