SelfSupervised SpatioTemporal Graph Learning for PointofInterest Recommendation
Abstract
:1. Introduction
 To the best of our knowledge, this is the first attempt to design a selfsupervised learningbased framework to improve GNNbased POI recommendation algorithms.
 We propose data augmentation strategies and pretext tasks of the proposed framework, which model spatial or temporal prior knowledge from different perspectives.
 We conducted experiments on three POI recommendation datasets and verified that our model could improve GNNbased POI recommendations and outperform existing stateoftheart methods.
2. Related Works
2.1. PointofInterest Recommendation
2.2. SelfSupervised Learning
3. Methodology
3.1. Problem Definition and GNN Backbone
3.2. SpatioTemporalAware Data Augmentation
 Spatial similarity matrix ${\mathcal{M}}_{S}\in {\{0,1\}}^{\left\mathcal{P}\right\times \left\mathcal{P}\right}$: when the distance between two POIs is less than a certain threshold ${K}_{S}$, then the similarity of these two POIs is 1; otherwise, it is 0.
 Temporal similarity matrix ${\mathcal{M}}_{T}\in {\{0,1\}}^{\left\mathcal{P}\right\times \left\mathcal{P}\right}$: when two POIs have interacted with the same user in a period ${K}_{T}$, then the similarity of these two POIs is 1; otherwise, it is 0.
 Spatialaware edge perturbation (SEP): It adds multiple implicit edges based on the spatial similarity to the original user–POI edges:$${\mathcal{E}}_{im{p}_{S}}^{\left\mathcal{U}\right\times \left\mathcal{P}\right}=\mathcal{E}\xb7{\mathcal{M}}_{S},$$$${s}_{1}\left(\mathcal{G}\right)=(\mathcal{V},{\mathbf{\Theta}}_{\mathbf{1}}({\mathcal{E}}_{im{p}_{S}},\mathcal{E})),\phantom{\rule{0.277778em}{0ex}}{s}_{2}\left(\mathcal{G}\right)=(\mathcal{V},{\mathbf{\Theta}}_{\mathbf{2}}({\mathcal{E}}_{im{p}_{S}},\mathcal{E})),$$
 Temporalaware edge perturbation (TEP): It adds multiple implicit edges based on temporal similarity to the original user–POI edges:$${\mathcal{E}}_{im{p}_{T}}^{\left\mathcal{U}\right\times \left\mathcal{P}\right}=\mathcal{E}\xb7{\mathcal{M}}_{T},$$$${s}_{1}\left(\mathcal{G}\right)=(\mathcal{V},{\mathbf{\Theta}}_{\mathbf{1}}({\mathcal{E}}_{im{p}_{T}},\mathcal{E})),\phantom{\rule{0.277778em}{0ex}}{s}_{2}\left(\mathcal{G}\right)=(\mathcal{V},{\mathbf{\Theta}}_{\mathbf{2}}({\mathcal{E}}_{im{p}_{T}},\mathcal{E})).$$
3.3. SpatioTemporalAware PreText Task
 Spatialaware pretext task (SPT): We took ${Q}_{S}$ nodes with the highest spatial similarity in ${\mathcal{M}}_{S}$ to the target node as positive examples and $\rho (\%)$ nodes with the lowest spatial similarity as negative examples.
 Temporalaware pretext task (TPT): We took ${Q}_{T}$ nodes with the highest temporal similarity in ${\mathcal{M}}_{T}$ to the target node as positive examples and $\rho (\%)$ nodes with the lowest temporal similarity as negative examples.
 Spatialaware contrastive learning (SCL): Maximizes the MI between spatialaware positive POI pairs and minimizes the MI between spatialaware negative POI pairs:$${\mathcal{L}}_{S}^{P}=\sum _{i\in \mathcal{P}}\mathrm{log}\frac{{\sum}_{j\in {Q}_{S}}\mathrm{exp}\left(s\left({\mathbf{z}}_{i}^{1},{\mathbf{z}}_{j}^{2}\right)/\tau \right)}{{\sum}_{k\in \left(\right\mathcal{P}\xb7{\rho}_{S})}\mathrm{exp}\left(s\left({\mathbf{z}}_{i}^{1},{\mathbf{z}}_{k}^{2}\right)/\tau \right)},$$
 Temporalaware contrastive learning (TCL): Maximizes the MI between temporalaware positive POI pairs and minimizes the MI between temporalaware negative POI pairs.$${\mathcal{L}}_{T}^{P}=\sum _{i\in \mathcal{P}}\mathrm{log}\frac{{\sum}_{j\in {Q}_{T}}\mathrm{exp}\left(s\left({\mathbf{z}}_{i}^{1},{\mathbf{z}}_{j}^{2}\right)/\tau \right)}{{\sum}_{k\in \left(\right\mathcal{P}\xb7{\rho}_{T})}\mathrm{exp}\left(s\left({\mathbf{z}}_{i}^{1},{\mathbf{z}}_{k}^{2}\right)/\tau \right)}.$$
3.4. Model Training
3.5. Complexity Analyses of SSTGL
Algorithm 1 The framework of SSTGL 
Require:
Given the original userPOI graph $\mathcal{G}=(\mathcal{V},\mathcal{E})$, the layer L of GNN model.
Ensure: The similarity scores ${\widehat{Y}}_{UP}$.

4. Experiments
 RQ1: How does the proposed SSTGL method perform when compared with the startoftheart baselines?
 RQ2: How does each component of the SSTGL contribute to the overall performance?
 RQ3: How do different hyperparameters influence the performance of SSTGL?
4.1. Experimental Setup
4.1.1. Datasets
 Foursquare [19]: The Foursquare dataset consists of checkin data generated on Foursquare from April 2012 to September 2013. Following [19], we removed users with less than 10 interactions and POIs with less than 10 interactions. After preprocessing, it contained 1,196,248 checkins between 24,941 users and 28,593 POIs.
 Gowalla [19]: The Gowalla dataset consists of checkin data generated on Gowalla from February 2009 to October 2010. As was done in [19], we removed users with less than 15 interactions and POIs with less than 10 interactions. After preprocessing, it contained 1,278,274 checkins between 18,737 users and 32,510 POIs.
 Meituan “https://www.biendata.xyz/competition/smp2021_1/ (accessed on 28 July 2023)”: The Meituan dataset consists of checkin data generated on the Meituan APP from 1st March 2021 to 28th March 2021. We removed users with less than 10 interactions and POIs without location information. After preprocessing, it contained 602,331 checkins between 38,904 users and 3182 POIs.
4.1.2. Baselines
 NeuMF [20]: NeuMF is a classical MFbased model that combines matrix factorization and multilayer perceptron to learn both lowdimensional and highdimensional embeddings.
 NGCF [21]: NGCF is a GNNbased model capturing highorder information through message passing and aggregation.
 DGCF [22]: DGCF is a GNNbased model, which models different relationships and separates user intents in the representation.
 LightGCN [23]: LightGCN is a GNNbased recommendation model, which simplifies the aggregation step by deleting the weight matrix and activation function.
 SGL [12]: SGL is a graphbased selfsupervised method that proposes three data augmentation strategies based on the graph structure.
 NCL [13]: NCL is a graphbased contrastive learning method that improves neural graph collaborative filtering by considering structural and semantic neighbors.
 LGLMF [3]: LGLMF is an MFbased POI recommendation model, which combines logistic matrix factorization with a regionbased geographical model.
 STACP [5]: STACP is also an MFbased POI recommendation model, which combines matrix factorization with a spatiotemporal activitycenters algorithm.
 GPR [9]: GPR is a GNNbased model designed for POI recommendation that uses an extra POI–POI graph to learn item embeddings and improve performance.
 MPGRec [24]: MPGRec is the newest GNNbased POI recommendation model, which uses a dynamic memory module to store global information for spatial consistency.
4.1.3. Evaluation Metrics
4.1.4. Implementation Details
4.2. Performance Comparison (RQ1)
 SSTGL(SEP): uses spatialaware edge perturbation and nonspatiotemporal pretext tasks.
 SSTGL(TEP): uses temporalaware edge perturbation and nonspatiotemporal pretext task.
 SSTGL(SCL): uses spatialaware contrastive learning and nonspatiotemporal data augmentation.
 SSTGL(TCL): use temporalaware contrastive learning and nonspatiotemporal data augmentation.
 SSTGL outperformed all baseline methods in most cases. In particular, the relative improvements from the strongest baselines were 6.32% (Foursquare), 13.27% (Gowalla), and 9.68% (Meituan) using the Recall@50 metric. Note that SSTGL not only worked better than the existing POI recommendation methods, but also better than the existing selfsupervised graph learning methods. This demonstrates the ability of our model to use selfsupervised learning to alleviate the data sparsity problem in the POI recommendation task. Although MPGRec performed better in some cases, it relies on a dynamic memory module, which requires a large memory overhead.
 For the baseline models, the GNN models did not always outperform the MF models, which was related to the datasets and model architectures. For example, we found that the NeuMF model performed better than some GNNbased methods for the Meituan dataset. This may be due to the low sparsity of the Meituan dataset and the more personalized interests of the users in the takeout scenario, so aggregating higherorder neighborhood information would instead reduce the performance.
4.3. Ablation Study (RQ2)
 In the strategies we designed, the temporalbased approaches (i.e., TEP and TCL) worked better on the Foursquare and Gowalla datasets, and the spatialaware approaches (i.e., SEP and SCL) performed better on the Meituan dataset. This may indicate that the spatial factor has a greater influence on the Meituan dataset compared with other datasets.
 Although both consider spatiotemporal information, the data augmentationbased approach outperformed the pretexttaskbased approach. This may be due to the direct modification of the graph structure using the selfsupervised method of data augmentation, which allows the node representation of the GNN output to make better use of spatiotemporal prior knowledge.
4.4. Influence of HyperParameters (RQ3)
 Overall, the different drop ratio ${r}_{\theta}$ and sample ratio $\rho $ had little effect on the model results, which indicates the robustness of the model.
 Too large or too small SSL temperature $\tau $ values reduced the performance. This observation is consistent with the previous work [12]. The possible reason behind this is that, if the temperature is large, it is more difficult to distinguish negative examples. If the temperature is small, only a small number of negative cases affect the optimization.
4.5. Discussion
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
 Werneck, H.; Silva, N.; Viana, M.; Pereira, A.C.; Mourão, F.; Rocha, L. Points of interest recommendations: Methods, evaluation, and future directions. Inf. Syst. 2021, 101, 101789. [Google Scholar] [CrossRef]
 Gao, H.; Tang, J.; Hu, X.; Liu, H. Exploring temporal effects for location recommendation on locationbased social networks. In Proceedings of the 7th ACM Conference on Recommender Systems, Hong Kong, China, 12–16 October 2013. [Google Scholar]
 Rahmani, H.A.; Aliannejadi, M.; Ahmadian, S.; Baratchi, M.; Afsharchi, M.; Crestani, F. LGLMF: Local geographical based logistic matrix factorization model for POI recommendation. In Proceedings of the Asia Information Retrieval Symposium, Hong Kong, China, 7–9 November 2019; Springer: Berlin/Heidelberg, Germany, 2019; pp. 66–78. [Google Scholar]
 Liu, Y.; Wei, W.; Sun, A.; Miao, C. Exploiting geographical neighborhood characteristics for location recommendation. In Proceedings of the 23rd ACM International Conference on Information and Knowledge Management, Shanghai, China, 3–7 November 2014. [Google Scholar]
 Rahmani, H.A.; Aliannejadi, M.; Baratchi, M.; Crestani, F. Joint geographical and temporal modeling based on matrix factorization for pointofinterest recommendation. In Proceedings of the European Conference on Information Retrieval, Lisbon, Portugal, 14–17 April 2020; Springer: Berlin/Heidelberg, Germany, 2020; pp. 205–219. [Google Scholar]
 Xie, M.; Yin, H.; Wang, H.; Xu, F.; Chen, W.; Wang, S. Learning graphbased poi embedding for locationbased recommendation. In Proceedings of the 25th ACM International on Conference on Information and Knowledge Management, Indianapolis, IN, USA, 24–28 October 2016. [Google Scholar]
 Yang, D.; Qu, B.; Yang, J.; CudreMauroux, P. Revisiting user mobility and social relationships in lbsns: A hypergraph embedding approach. In Proceedings of the World Wide Web Conference, San Francisco, CA, USA, 13–17 May 2019. [Google Scholar]
 Qian, T.; Liu, B.; Nguyen, Q.V.H.; Yin, H. Spatiotemporal representation learning for translationbased POI recommendation. ACM Trans. Inf. Syst. 2019, 37, 18. [Google Scholar] [CrossRef]
 Chang, B.; Jang, G.; Kim, S.; Kang, J. Learning graphbased geographical latent representation for pointofinterest recommendation. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management, Virtual, 19–23 October 2020. [Google Scholar]
 Zhong, T.; Zhang, S.; Zhou, F.; Zhang, K.; Trajcevski, G.; Wu, J. Hybrid graph convolutional networks with multihead attention for location recommendation. World Wide Web 2020, 23, 3125–3151. [Google Scholar] [CrossRef]
 Li, Z.; Cheng, W.; Xiao, H.; Yu, W.; Chen, H.; Wang, W. You Are What and Where You Are: Graph Enhanced Attention Network for Explainable POI Recommendation. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management, Virtual, 1–5 November 2021; pp. 3945–3954. [Google Scholar]
 Wu, J.; Wang, X.; Feng, F.; He, X.; Chen, L.; Lian, J.; Xie, X. Selfsupervised graph learning for recommendation. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual, 11–15 July 2021; pp. 726–735. [Google Scholar]
 Lin, Z.; Tian, C.; Hou, Y.; Zhao, W.X. Improving Graph Collaborative Filtering with Neighborhoodenriched Contrastive Learning. In Proceedings of the ACM Web Conference 2022, Lyon, France, 25–29 April 2022. [Google Scholar]
 Yu, J.; Yin, H.; Li, J.; Wang, Q.; Hung, N.Q.V.; Zhang, X. Selfsupervised multichannel hypergraph convolutional network for social recommendation. In Proceedings of the Web Conference 2021, Ljubljana, Slovenia, 19–23 April 2021; pp. 413–424. [Google Scholar]
 Xia, X.; Yin, H.; Yu, J.; Wang, Q.; Cui, L.; Zhang, X. Selfsupervised hypergraph convolutional networks for sessionbased recommendation. In Proceedings of the AAAI Conference on Artificial Intelligence, Virtual, 2–9 February 2021. [Google Scholar]
 He, K.; Fan, H.; Wu, Y.; Xie, S.; Girshick, R. Momentum contrast for unsupervised visual representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 13–19 June 2020. [Google Scholar]
 Devlin, J.; Chang, M.W.; Lee, K.; Toutanova, K. Bert: Pretraining of deep bidirectional transformers for language understanding. arXiv 2018, arXiv:1810.04805. [Google Scholar]
 Han, H.; Zhang, M.; Hou, M.; Zhang, F.; Wang, Z.; Chen, E.; Wang, H.; Ma, J.; Liu, Q. STGCN: A spatialtemporal aware graph learning method for POI recommendation. In Proceedings of the 2020 IEEE International Conference on Data Mining (ICDM), Sorrento, Italy, 17–20 November 2020. [Google Scholar]
 Liu, Y.; Pham, T.A.N.; Cong, G.; Yuan, Q. An experimental evaluation of pointofinterest recommendation in locationbased social networks. Proc. VLDB Endow. 2017, 10, 1010–1021. [Google Scholar] [CrossRef]
 He, X.; Liao, L.; Zhang, H.; Nie, L.; Hu, X.; Chua, T.S. Neural collaborative filtering. In Proceedings of the 26th International Conference on World Wide Web, Perth, Australia, 3–7 April 2017; pp. 173–182. [Google Scholar]
 Wang, X.; He, X.; Wang, M.; Feng, F.; Chua, T.S. Neural graph collaborative filtering. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris, France, 21–25 July 2019. [Google Scholar]
 Wang, X.; Jin, H.; Zhang, A.; He, X.; Xu, T.; Chua, T.S. Disentangled graph collaborative filtering. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual, 25–30 July 2020; pp. 1001–1010. [Google Scholar]
 He, X.; Deng, K.; Wang, X.; Li, Y.; Zhang, Y.; Wang, M. Lightgcn: Simplifying and powering graph convolution network for recommendation. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual, 25–30 July 2020. [Google Scholar]
 Yang, T.; Gao, H.; Yang, C.; Shi, C.; Xie, Q.; Wang, X.; Wang, D. MemoryEnhanced PeriodAware Graph Neural Network for General POI Recommendation. In Proceedings of the International Conference on Database Systems for Advanced Applications, Tianjin, China, 17–20 April 2023; Springer: Berlin/Heidelberg, Germany, 2023; pp. 462–472. [Google Scholar]
 Acharya, M.; Yadav, S.; Mohbey, K.K. How can we create a recommender system for tourism? A location centric spatial binningbased methodology using social networks. Int. J. Inf. Manag. Data Insights 2023, 3, 100161. [Google Scholar] [CrossRef]
 Acharya, M.; Mohbey, K.K. Differential PrivacyBased Social Network Detection Over SpatioTemporal Proximity for Secure POI Recommendation. SN Comput. Sci. 2023, 4, 252. [Google Scholar] [CrossRef]
 Lim, N.; Hooi, B.; Ng, S.K.; Wang, X.; Goh, Y.L.; Weng, R.; Varadarajan, J. STPUDGAT: Spatialtemporalpreference user dimensional graph attention network for next POI recommendation. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management, Online, 19–23 October 2020. [Google Scholar]
 Li, Y.; Chen, T.; Yin, H.; Huang, Z. Discovering collaborative signals for next POI recommendation with iterative Seq2Graph augmentation. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence (IJCAI21), Virtual, 19–26 August 2021. [Google Scholar]
 Zhou, F.; Wang, P.; Xu, X.; Tai, W.; Trajcevski, G. Contrastive trajectory learning for tour recommendation. In ACM Transactions on Intelligent Systems and Technology (TIST); Association for Computing Machinery: New York, NY, USA, 2021. [Google Scholar]
 Zhao, W.X.; Mu, S.; Hou, Y.; Lin, Z.; Chen, Y.; Pan, X.; Li, K.; Lu, Y.; Wang, H.; Tian, C.; et al. Recbole: Towards a unified, comprehensive and efficient framework for recommendation algorithms. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management, Online, 1–5 November 2021. [Google Scholar]
 Kingma, D.P.; Ba, J. Adam: A Method for Stochastic Optimization. In Proceedings of the 3rd International Conference on Learning Representations, ICLR 2015, Diego, CA, USA, 7–9 May 2015. [Google Scholar]
Dataset  #CheckIns  #POI  #User  Sparsity  Time Span 

Foursquare  1,196,248  28,593  24,941  99.90%  April 2012–September 2013 
Gowalla  1,278,274  32,510  18,737  99.87%  February 2009–October 2010 
Meituan  602,331  3182  38,904  99.51%  1st March 2021–28th March 2021 
MFBased  GNNBased  GNN and SSLBased  

STunaware  NeuMF [20]  NGCF [21], DGCF [22], LightGCN [23]  SGL [12], NCL [13] 
STaware  LGLMF [3], STACP [5]  GPR [9], MPGRec [24]  SSTGL(Ours) 
Model  Recall@5  Recall@10  Recall@20  Recall@50  MAP@5  MAP@10  MAP@20  MAP@50 

NeuMF  0.0368  0.0610  0.0981  0.1727  0.0235  0.0241  0.0271  0.0310 
NGCF  0.0390  0.0627  0.0980  0.1688  0.0249  0.0250  0.0278  0.0314 
DGCF  0.0435  0.0669  0.1028  0.1764  0.0291  0.0288  0.0316  0.0354 
LightGCN  0.0469  0.0721  0.1076  0.1796  0.0317  0.0312  0.0341  0.0378 
SGL  0.0452  0.0707  0.1080  0.1852  0.0300  0.0299  0.0330  0.0371 
NCL  0.0463  0.0723  0.1083  0.1839  0.0313  0.0310  0.0338  0.0378 
STACP  0.0274  0.0450  0.0700  0.1275  0.0187  0.0186  0.0206  0.0235 
LGLMF  0.0284  0.0459  0.0729  0.1284  0.0192  0.0190  0.0212  0.0242 
GPR  0.0316  0.0502  0.0763  0.1272  0.0183  0.0205  0.0224  0.0243 
MPGRec  0.0592  0.0848  0.1200  0.1915  0.0366  0.0398  0.0425  0.0452 
SSTGL (Ours)  0.0577  0.0851  0.1244  0.2036  0.0338  0.0374  0.0401  0.0427 
Model  Recall@5  Recall@10  Recall@20  Recall@50  Map@5  Map@10  Map@20  Map@50 

NeuMF  0.0302  0.0497  0.0808  0.1478  0.0227  0.0211  0.0231  0.0267 
NGCF  0.0308  0.0500  0.0810  0.1458  0.0235  0.0216  0.0234  0.0268 
DGCF  0.0332  0.0530  0.0834  0.1477  0.0266  0.0239  0.0254  0.0288 
LightGCN  0.0352  0.0564  0.0897  0.1593  0.0271  0.0247  0.0267  0.0305 
SGL  0.0338  0.0557  0.0911  0.1657  0.0259  0.0240  0.0262  0.0305 
NCL  0.0344  0.0561  0.0902  0.1631  0.0267  0.0244  0.0264  0.0305 
STACP  0.0176  0.0302  0.0509  0.0964  0.0142  0.0131  0.0143  0.0168 
LGLMF  0.0241  0.0398  0.0646  0.1156  0.0209  0.0188  0.0201  0.0230 
GPR  0.0302  0.0483  0.0766  0.1310  0.0183  0.0196  0.0216  0.0238 
MPGRec  0.0471  0.0706  0.1050  0.1718  0.0308  0.0325  0.0349  0.0377 
SSTGL (Ours)  0.0511  0.0786  0.1175  0.1946  0.0292  0.0329  0.0357  0.0383 
Model  Recall@5  Recall@10  Recall@20  Recall@50  MAP@5  MAP@10  MAP@20  MAP@50 

NeuMF  0.3540  0.4315  0.4631  0.5063  0.2339  0.2517  0.2562  0.2588 
NGCF  0.3198  0.3952  0.4499  0.5177  0.2123  0.228  0.2349  0.2392 
DGCF  0.3285  0.4046  0.4612  0.5335  0.2194  0.2353  0.2427  0.2472 
LightGCN  0.3456  0.4221  0.4705  0.5315  0.2250  0.2417  0.2482  0.252 
SGL  0.3373  0.4123  0.4674  0.5311  0.2432  0.2591  0.2664  0.2705 
NCL  0.3466  0.4236  0.4667  0.5187  0.2222  0.2391  0.2451  0.2484 
STACP  0.0054  0.0094  0.0203  0.0421  0.0022  0.0026  0.0034  0.0041 
LGLMF  0.0005  0.0010  0.0030  0.0087  0.0002  0.0002  0.0004  0.0005 
GPR  0.2984  0.3758  0.4573  0.5669  0.2066  0.2185  0.2253  0.2297 
MPGRec  0.3930  0.4316  0.4655  0.5147  0.3170  0.3240  0.3272  0.3294 
SSTGL (Ours)  0.3865  0.4388  0.4917  0.5645  0.3073  0.3146  0.3185  0.3209 
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. 
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Liu, J.; Gao, H.; Shi, C.; Cheng, H.; Xie, Q. SelfSupervised SpatioTemporal Graph Learning for PointofInterest Recommendation. Appl. Sci. 2023, 13, 8885. https://doi.org/10.3390/app13158885
Liu J, Gao H, Shi C, Cheng H, Xie Q. SelfSupervised SpatioTemporal Graph Learning for PointofInterest Recommendation. Applied Sciences. 2023; 13(15):8885. https://doi.org/10.3390/app13158885
Chicago/Turabian StyleLiu, Jiawei, Haihan Gao, Chuan Shi, Hongtao Cheng, and Qianlong Xie. 2023. "SelfSupervised SpatioTemporal Graph Learning for PointofInterest Recommendation" Applied Sciences 13, no. 15: 8885. https://doi.org/10.3390/app13158885