Next point of interest recommendation based on spatio-temporal causal inference

doi:10.11947/j.AGCS.2025.20250202

Abstract

Abstract:

The next point of interest (POI) recommendation is a vital service within location-based social networks, demonstrating substantial application potential in domains such as personalized location recommendation for users and layout optimization for business. However, current recommendation systems predominantly leverage deep learning approaches without incorporating causal inference frameworks. This limitation results in a tendency to learn homogenized correlations, which renders the systems incapable of effectively mitigating the influence from confounding factors inherent in spatio-temporal data. Consequently, the recommendation performance of these models is significantly constrained. To address the mentioned issues, this paper proposes spatio-temporal causal inference network (STCIN), a next POI recommendation method based on spatio-temporal causal inference. First, to integrate spatio-temporal data into the causal inference framework, we design a spatio-temporal correlation embedding module. This module separately encodes sequential information and spatio-temporal contextual information through temporal feature-based and spatial feature-based embedding operations, generating user and POI feature representations from a spatio-temporal perspective. Next, we propose a causal-inference module based on front-door criterion and counterfactual reasoning. It discerns the causal effects among feature variables, and distills users' spatio-temporal states that capture individual heterogeneity, to alleviate the influence of confounding factors within spatio-temporal data. Finally, the model processes multi-period spatio-temporal states, estimating the probability that a user will visit each candidate POI, to achieve next POI recommendation. Extensive experiments conducted on two real-world datasets demonstrate that, compared to the best baseline model, the STCIN improves accuracy and mean reciprocal rank by 37.60% and 22.72% on New York dataset, and by 32.84% and 20.63% on Tokyo dataset. These results substantiate the effectiveness and superiority of the proposed STCIN model.

Key words: next POI recommendation, spatio-temporal context information, causal inference, confounding factor

CLC Number:

P208

Xinyu YE, Shenghua XU, Jiping LIU, Hongyu CHEN, Zhuolu WANG, Weilian LI. Next point of interest recommendation based on spatio-temporal causal inference[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(11): 2081-2096.

Figures/Tables 17

Fig. 1

Tab. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Fig. 8

Fig. 9

Tab. 2

Tab. 3

Tab. 4

Tab. 5

Tab. 6

Fig. 10

Fig. 11

References 55

[1]	王勇, 马钰, 徐胜华, 等. 兴趣点推荐方法研究进展与展望[J]. 测绘科学, 2023, 48(12): 217-224.
	WANG Yong, MA Yu, XU Shenghua, et al. Research progress and prospects of point of interest recommendation methods[J]. Science of Surveying and Mapping, 2023, 48(12): 217-224.
[2]	郭旦怀, 张鸣珂, 贾楠, 等. 融合深度学习技术的用户兴趣点推荐研究综述[J]. 武汉大学学报(信息科学版), 2020, 45(12): 1890-1902.
	GUO Danhuai, ZHANG Mingke, JIA Nan, et al. Survey of point-of-interest recommendation research fused with deep learning[J]. Geomatics and Information Science of Wuhan University, 2020, 45(12): 1890-1902.
[3]	ZHOU Zhengyang, YANG Kuo, LIANG Yuxuan, et al. Predicting collective human mobility via countering spatiotemporal heterogeneity[J]. IEEE Transactions on Mobile Computing, 2024, 23(5): 4723-4738.
[4]	YAO Di, ZHANG Chao, HUANG Jianhui, et al. SERM: a recurrent model for next location prediction in semantic trajectories[C]//Proceedings of 2017 ACM on Conference on Information and Knowledge Management. Singapore: ACM Press, 2017: 2411-2414.
[5]	SÁNCHEZ P, BELLOGÍN A. Point-of-interest recommender systems based on location-based social networks: a survey from an experimental perspective[J]. ACM Computing Surveys, 2022, 54(11s): 1-37.
[6]	YE Ziming, ZHANG Xiao, CHEN Xu, et al. Adaptive clustering based personalized federated learning framework for next POI recommendation with location noise[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 36(5): 1843-1856.
[7]	SARKAR J L, MAJUMDER A, PANIGRAHI C R, et al. MULTITOUR: a multiple itinerary tourists recommendation engine[J]. Electronic Commerce Research and Applications, 2020, 40: 100943.
[8]	PANG Guangyao, WANG Xiaoming, HAO Fei, et al. Efficient point-of-interest recommendation with hierarchical attention mechanism[J]. Applied Soft Computing, 2020, 96: 106536.
[9]	李曼文, 张月琴, 张晨威, 等. 异质图嵌入的地理不敏感时空兴趣点推荐方法[J]. 计算机科学与探索, 2024, 18(3): 755-767.
	LI Manwen, ZHANG Yueqin, ZHANG Chenwei, et al. Geographically insensitive spatial-temporal POI recommendation based on heterogeneous graph embedding[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(3): 755-767.
[10]	孟祥武, 李瑞昌, 张玉洁, 等. 基于用户轨迹数据的移动推荐系统研究[J]. 软件学报, 2018, 29(10): 3111-3133.
	MENG Xiangwu, LI Ruichang, ZHANG Yujie, et al. Survey on mobile recommender systems based on user trajectory data[J]. Journal of Software, 2018, 29(10): 3111-3133.
[11]	李鹏飞, 贺洋, 毋建宏. 融合全局特征的时空网络兴趣点推荐算法[J]. 计算机工程与应用, 2024, 60(11): 75-83.
	LI Pengfei, HE Yang, WU Jianhong. Spatio-temporal network interest point recommendation algorithm fusing global features[J]. Computer Engineering and Applications, 2024, 60(11): 75-83.
[12]	ZHAO Shengjin, ZHAO Tong, YANG Haiqin, et al. STELLAR: spatial-temporal latent ranking for successive point-of-interest recommendation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2016, 30(1): 315-321.
[13]	LIU Qiang, WU Shu, WANG Liang, et al. Predicting the next location: a recurrent model with spatial and temporal contexts[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2016, 30(1): 194-200.
[14]	LI Xixi, HU Ruimin, WANG Zheng. Beyond fixed time and space: next POI recommendationvia multi-grained context and correlation[J]. Neural Computing and Applications, 2023, 35(1): 907-920.
[15]	ACHARYA M, MOHBEY K K. Exploring the evolution, progress, and future of point-of-interest recommendation over location-based social network: a comprehensive review[J]. GeoInformatica, 2025, 29(3): 305-350.
[16]	刘广腾, 王峰, 吴中博. 下一个兴趣点推荐算法综述[J]. 计算机科学与探索, 2025, 19(7): 1747-1770.
	LIU Guangteng, WANG Feng, WU Zhongbo. Survey on next point-of-interest recommendation algorithms[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(7): 1747-1770.
[17]	ZHAO Pengfei. Optimization of LSTM ship trajectory prediction based on hybrid genetic algorithm[J]. Journal of Geodesy and Geoinformation Science, 2024, 7(3): 89-102.
[18]	ZHAO Pengpeng, ZHU Haifeng, LIU Yanchi, et al. Where to go next: a spatio-temporal gated network for next POI recommendation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 5877-5884.
[19]	LAI Yantong, SU Yijun, WEI Lingwei, et al. Adaptive spatial-temporal hypergraph fusion learning for next POI recommendation[C]//Proceedings of 2024 IEEE International Conference on Acoustics, Speech and Signal Processing. Seoul: IEEE, 2024: 7320-7324.
[20]	DAI Shaojie, YU Yanwei, FAN Hao, et al. Spatio-temporal representation learning with social tie for personalized POI recommendation[J]. Data Science and Engineering, 2022, 7(1): 44-56.
[21]	FENG Shanshan, LI Xutao, ZENG Yifeng, et al. Personalized ranking metric embedding for next new POI recommendation[C]//Proceedings of the 24th International Conference on Artificial Intelligence. [S.l.]: IACM Press, 2015: 2069-2075.
[22]	XU Xiaohang, SUZUMURA T, YONG Jiawei, et al. Revisiting mobility modeling with graph: a graph transformer model for next point-of-interest recommendation[C]//Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems. Hamburg: ACM Press, 2023: 1-10.
[23]	SONG Weiping, XIAO Zhiping, WANG Yifan, et al. Session-based social recommendation via dynamic graph attention networks[C]//Proceedings of the 12th ACM International Conference on Web Search and Data Mining. Melbourne: ACM Press, 2019: 555-563.
[24]	YAO Liuyi, CHU Zhixuan, LI Sheng, et al. A survey on causal inference[J]. ACM Transactions on Knowledge Discovery from Data, 2021, 15(5): 41-46.
[25]	DE HAAN P, JAYARAMAN D, LEVINE S. Causal confusion in imitation learning[J]. Advances in Neural Information Processing Systems, 2019, 32: 451-458.
[26]	DENG Pan, ZHAO Yu, LIU Junting, et al. Spatio-temporal neural structural causal models for bike flow prediction[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(4): 4242-4249.
[27]	杨新新, 刘真, 卢思博, 等. 基于因果推断的推荐系统去偏研究综述[J]. 计算机学报, 2024, 47(10): 2307-2332.
	YANG Xinxin, LIU Zhen, LU Sibo, et al. A survey on debiasing recommendation based on causal inference[J]. Chinese Journal of Computers, 2024, 47(10): 2307-2332.
[28]	ZHANG Yang, FENG Fuli, HE Xiangnan, et al. Causal intervention for leveraging popularity bias in recommendation[C]//Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. [S.l.]: ACM Press, 2021: 11-20.
[29]	THOMAS C K, CHACCOUR C, SAAD W, et al. Causal reasoning: charting a revolutionary course for next-generation AI-native wireless networks[J]. IEEE Vehicular Technology Magazine, 2024, 19(1): 16-31.
[30]	SUN Guohao, HUA Huirong, LU Jinhu, et al. A novel causal discovery model for recommendation system[M]//Web and Big Data. Singapore: Springer, 2024: 261-276.
[31]	GAO Chongming, WANG Shiqi, LI Shijun, et al. CIRS: bursting filter bubbles by counterfactual interactive recommender system[J]. ACM Transactions on Information Systems, 2024, 42(1): 1-27.
[32]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017(30): 151-163.
[33]	QIN Yanjun, FANG Yuchen, LUO Haiyong, et al. Next point-of-interest recommendation with auto-correlation enhanced multi-modal transformer network[C]//Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. Madrid: ACM Press, 2022: 2612-2616.
[34]	YANG Song, LIU Jiamou, ZHAO Kaiqi. GETNext: trajectory flow map enhanced transformer for next POI recommendation[C]//Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. Madrid: ACM Press, 2022: 1144-1153.
[35]	WANG Xinfeng, FUKUMOTO F, LI Jiyi, et al. STaTRL: Spatial-temporal and text representation learning for POI recommendation[J]. Applied Intelligence, 2023, 53(7): 8286-8301.
[36]	XU Shuqiang, HUANG Qunying, ZOU Zhiqiang. Spatio-temporal transformer recommender: next location recommendation with attention mechanism by mining the spatio-temporal relationship between visited locations[J]. ISPRS International Journal of Geo-Information, 2023, 12(2): 79.
[37]	HE Yuhang, ZHOU Wei, LUO Fengji, et al. Feature-based POI grouping with transformer for next point of interest recommendation[J]. Applied Soft Computing, 2023, 147: 110754.
[38]	KONG Xiangjie, CHEN Zhiyu, LI Jianxin, et al. KGNext: knowledge-graph-enhanced transformer for next POI recommendation with uncertain check-ins[J]. IEEE Transactions on Computational Social Systems, 2024, 11(5): 6637-6648.
[39]	SCARSELLI F, GORI M, TSOI A C, et al. The graph neural network model[J]. IEEE Transactions on Neural Networks, 2009, 20(1): 61-80.
[40]	KIM J, JEONG S, PARK G, et al. DynaPosGNN: dynamic-positional GNN for next POI recommendation[C]//Proceedings of 2021 International Conference on Data Mining Workshops. Auckland: IEEE, 2021: 36-44.
[41]	MENG Lingqiang, LIU Zhizhong, CHU Dianhui, et al. POI recommendation for occasional groups based on hybrid graph neural networks[J]. Expert Systems with Applications, 2024, 237: 121583.
[42]	FANG Jinfeng, MENG Xiangfu, QI Xueyue. A top-k POI recommendation approach based on LBSN and multi-graph fusion[J]. Neurocomputing, 2023, 518: 219-230.
[43]	WANG Zhaobo, ZHU Yanmin, WANG Chunyang, et al. Adaptive graph representation learning for next POI recommendation[C]//Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. Taipei: ACM Press, 2023: 393-402.
[44]	LIU Jiawei, GAO Haihan, YANG Cheng, et al. Heterogeneous spatio-temporal graph contrastive learning for point-of-interest recommendation[J]. Tsinghua Science and Technology, 2025, 30(1): 186-197.
[45]	ZHANG Dong, ZHANG Hanwang, TANG Jinhui, et al. Causal intervention for weakly-supervised semantic segmentation[J]. Advances in Neural Information Processing Systems, 2020, 33: 655-666.
[46]	SCHöLKOPF B. Causality for machine learning[DB/OL]. [2025-09-02]. https://dl.acm.org/doi/abs/10.1145/3501714.3501755.
[47]	LIN Xiangru, CHEN Yuyang, LI Guanbin, et al. A causal inference look at unsupervised video anomaly detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36(2): 1620-1629.
[48]	HE Ming, CHEN Xin, HU Xinlei, et al. Causal intervention for sentiment de-biasing in recommendation[C]//Proceedings of the 31st ACM International Conference on Information & Knowledge Management. Atlanta: ACM Press, 2022: 4014-4018.
[49]	LIU Bo, ZENG Jun, WEN Junhao, et al. CBRec: a causal way balancing multidimensional attraction effect in POI recommendations[J]. Knowledge-Based Systems, 2024, 305: 112607.
[50]	YANG Mengyue, DAI Quanyu, DONG Zhenhua, et al. Top-N recommendation with counterfactual user preference simulation[C]//Proceedings of the 30th ACM International Conference on Information & Knowledge Management. Queensland: ACM Press, 2021: 2342-2351.
[51]	RENDLE S, FREUDENTHALER C, SCHMIDT-THIEME L. Factorizing personalized Markov chains for next-basket recommendation[C]//Proceedings of the 19th International Conference on World Wide Web. Raleigh: ACM Press, 2010: 811-820.
[52]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
[53]	ZHANG Lu, SUN Zhu, WU Ziqing, et al. Next point-of-interest recommendation with inferring multi-step future preferences[C]//Proceedings of the 31th International Joint Conference on Artificial Intelligence. Vienna: [s.n.], 2022: 3751-3757.
[54]	LUO Yingtao, LIU Qiang, LIU Zhaocheng. STAN: spatio-temporal attention network for next location recommendation[C]//Proceedings of 2021 Web Conference. Ljubljana: ACM Press, 2021: 2177-2185.
[55]	HUANG Tianhao, PAN Xuan, CAI Xiangrui, et al. Learning time slot preferences via mobility tree for next POI recommendation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(8): 8535-8543.

符号	描述
	用户签到记录
	时空动态特征
	用户偏好特征
	上下文特征
	用户时空状态
	预测目标

用户	POI	POI类别	POI类别名称	签到时间戳	经度	纬度
71	481dfa3af964a5207f4f1fe3	4bf58dd8d48988d1fa931735	Hotel（酒店）	2012-04-03	73.99°W	40.75°N
71	481dfa3af964a5207f4f1fe3	4bf58dd8d48988d1fa931735	Hotel（酒店）	20：08：03	73.99°W	40.75°N
65	4d66b1ef84f28cfa548a6e69	4bf58dd8d48988d153941735	Burrito Place（卷饼店）	2012-04-03	73.99°W	40.81°N
65	4d66b1ef84f28cfa548a6e69	4bf58dd8d48988d153941735	Burrito Place（卷饼店）	20：09：55	73.99°W	40.81°N
592	4b44a049f964a5201cf825e3	4bf58dd8d48988d1fd931735	Subway（地铁）	2012-04-03	74.01°W	40.71°N
592	4b44a049f964a5201cf825e3	4bf58dd8d48988d1fd931735	Subway（地铁）	20：11：13	74.01°W	40.71°N
267	4e2bada17d8b7deda6d3fda6	4f2a25ac4b909258e854f55f	Neighborhood（街区）	2012-04-03	73.83°W	40.87°N
267	4e2bada17d8b7deda6d3fda6	4f2a25ac4b909258e854f55f	Neighborhood（街区）	20：11：28	73.83°W	40.87°N
976	4b5ca9dcf964a520f83c29e3	4bf58dd8d48988d1fd931735	Subway（地铁）	2012-04-03	73.93°W	40.70°N
976	4b5ca9dcf964a520f83c29e3	4bf58dd8d48988d1fd931735	Subway（地铁）	20：11：45	73.93°W	40.70°N

数据集	用户数	POI类别数	POI数	签到数	轨迹数
NYC数据集	1075	318	5099	104 074	14 160
TKY数据集	2281	291	7844	361 430	44 692

模型	NYC数据集				TKY数据集
模型	Acc@1	Acc@5	Acc@10	MRR	Acc@1	Acc@5	Acc@10	MRR
FPMC	0.100 3	0.212 6	0.297 0	0.170 1	0.081 4	0.204 5	0.274 6	0.134 4
PRME	0.115 9	0.223 6	0.310 5	0.171 2	0.105 2	0.272 8	0.294 4	0.178 6
LSTM	0.130 5	0.271 9	0.328 3	0.185 7	0.133 5	0.272 8	0.327 7	0.183 4
ST-RNN	0.148 3	0.292 3	0.362 2	0.219 8	0.140 9	0.302 2	0.357 7	0.221 2
STGN	0.171 6	0.338 1	0.412 2	0.259 8	0.168 9	0.339 1	0.384 8	0.242 2
CFPRec	0.169 2	0.386 7	0.489 4	0.268 0	0.205 2	0.402 8	0.476 9	0.296 3
STAN	0.223 1	0.458 2	0.573 4	0.325 3	0.196 3	0.379 8	0.446 4	0.285 2
GETNext	0.243 5	0.508 9	0.614 3	0.362 1	0.225 4	0.441 7	0.528 7	0.326 2
MTNet	0.262 0	0.538 1	0.632 1	0.385 5	0.257 5	0.497 7	0.584 8	0.365 9
STCIN	0.360 5	0.603 3	0.701 0	0.473 1	0.342 1	0.568 3	0.658 0	0.441 4
提升/（%）	37.60	12.12	10.90	22.72	32.84	14.18	12.52	20.63

试验组名称	时空关联嵌入	循环神经网络	图神经网络	反事实推理增强
STCIN	√	√	√	√
STCIN-P	√	×	√	√
STCIN-GEO	√	√	×	√
STCIN-R	√	√	√	×
STCIN-C	√	×	×	×