1. College of Computer Science and Technology, Heilongjiang University, Harbin 150080, China 2. Key Laboratory of Database and Parallel Computing, Heilongjiang University, Harbin 150080, China 3. Shandong Artificial Intelligence Institute, Qilu University of Technology, Jinan 250014, China 4. School of Mathematics and Statistics, Qilu University of Technology, Jinan 250353, China
Traditional collaborative filtering-based point-of-interest (POI) recommendation methods suffer from data sparsity and existing work tends to use contextual information alone without a reasonable balance of the effect of each factor. A point-of-interest recommendation model integrating user perception and multi-factor (UPMF) was proposed. User similarities for user perception-based implicit modelling were extracted to enrich user representations, and contextual information such as sequence, geography and social was used to construct POI recommendation models with synergistic user perception influences, which alleviated data sparsity. A novel user perception integration strategy (UPIS) was designed to mine the dynamic preferences of users based on user perceptions while making reasonable use of various contextual information. In addition, a segmentation-based activity area selection algorithm was proposed to model the impact of different activity areas on users. Results show that UPMF can improve the performance on metrics of precision, recall and normalized discounted cumulative gain (NDCG) compared to other popular POI recommendation technologies. Specifically, UPMF outperforms SUCP by 12.77% and 7.24% on the Gowalla and Yelp datasets in terms of NDCG@10, respectively.
Fig.1Framework diagram of point-of-interest recommendations model integrating user perception and multi-factor (UPMF)
数据集
$ \left| U \right| $
$ \left| L \right| $
$ \left| S \right| $
spa/%
Gowalla
5628
31803
620683
99.65
Yelp
7135
16621
301753
99.75
Tab.1Statistical information about dataset
对比模型
Gowalla
Yelp
P@10
P@20
R@10
R@20
NDCG@10
NDCG@20
P@10
P@20
R@10
R@20
NDCG@10
NDCG@20
TopPopular
0.0192
0.0146
0.0176
0.0270
0.0088
0.0079
0.0077
0.0073
0.0095
0.0185
0.0075
0.0073
PFMMGM
0.0240
0.0207
0.0258
0.0442
0.0140
0.0144
0.0162
0.0135
0.0242
0.0402
0.0175
0.0152
LMFT
0.0315
0.0269
0.0303
0.0515
0.0157
0.0150
0.0183
0.0163
0.0264
0.0457
0.0194
0.0176
iGLSR
0.0297
0.0242
0.0283
0.0441
0.0153
0.0145
0.0235
0.0192
0.0306
0.0529
0.0231
0.0214
GeoSoCa
0.0215
0.0195
0.0253
0.0449
0.0222
0.0206
0.0183
0.0150
0.0214
0.0340
0.0195
0.0168
MLP
0.0243
0.0215
0.0237
0.0396
0.0098
0.0127
0.0203
0.0174
0.0284
0.0157
0.0195
0.0185
Rank-GeoFM
0.0352
0.0297
0.0379
0.0602
0.0187
0.0179
0.0231
0.0198
0.0316
0.0587
0.0217
0.0214
L-WMF
0.0341
0.0296
0.0351
0.0582
0.0183
0.0178
0.0215
0.0181
0.0295
0.0542
0.0202
0.0194
LFBCA
0.0453
0.0376
0.0460
0.0742
0.0492
0.0427
0.0269
0.0228
0.0408
0.0667
0.0290
0.0255
CARA
0.0501
0.0410
0.0485
0.0792
0.0531
0.0452
0.0255
0.0219
0.0362
0.0628
0.0271
0.0239
SUCP
0.0502
0.0410
0.0520
0.0823
0.0548
0.0470
0.0286
0.0245
0.0438
0.0731
0.0304
0.0270
UPMF
0.0556
0.0445
0.0571
0.0886
0.0618
0.0519
0.0307
0.0267
0.0460
0.0784
0.0326
0.0292
Imp/%
10.76
8.53
9.81
7.65
12.77
10.43
7.34
8.98
5.02
7.25
7.24
8.15
Tab.2Performance comparison of P@N, R@N and NDCG@N between UPMF model and baselines on Gowalla and Yelp datasets
对比模型
Gowalla
Yelp
P@10
P@20
R@10
R@20
NDCG@10
NDCG@20
P@10
P@20
R@10
R@20
NDCG@10
NDCG@20
UPMF-sim
0.0555
0.0439
0.0556
0.0860
0.0609
0.0512
0.0300
0.0258
0.0444
0.0765
0.0316
0.0282
UPMF-SGM
0.0444
0.0360
0.0438
0.0672
0.0480
0.0411
0.0223
0.0201
0.0320
0.0568
0.0234
0.0215
UPMF-GM
0.0539
0.0442
0.0538
0.0865
0.0593
0.0509
0.0300
0.0257
0.0454
0.0764
0.0321
0.0285
UPMF-SM
0.0516
0.0422
0.0519
0.0833
0.0566
0.0486
0.0289
0.0252
0.0440
0.0746
0.0311
0.0278
UPMF-UPIS
0.0550
0.0441
0.0553
0.0866
0.0598
0.0508
0.0306
0.0258
0.0470
0.0774
0.0326
0.0287
UPMF
0.0556
0.0445
0.0571
0.0886
0.0618
0.0519
0.0307
0.0267
0.0460
0.0784
0.0326
0.0292
Tab.3Performance comparison of P@N、R@N and NDCG@N between UPMF model and its variant models on Gowalla and Yelp datasets
Fig.2Effect of data sparsity on Recall@20 and NDCG@20 on datasets Gowalla and Yelp
Fig.3Effect of different parameters of user perception integration strategy (UPIS)
[1]
HU Y, KOREN Y, VOLINSKY C. Collaborative filtering for implicit feedback datasets[C]// 8th IEEE International Conference on Data Mining. Pisa: IEEE, 2008: 263-272.
[2]
LIU K, ZHENG W, XIAO Y, et al. POI recommendation algorithm based on region transfer collaborative filtering [C]// IEEE 25th International Conference on Computer Supported Cooperative Work in Design (CSCWD). Hangzhou: IEEE, 2022: 903-907.
[3]
MNIH A, SALAKHUTDINOV R. Probabilistic matrix factorization[C]// Advances in Neural Information Processing Systems. Vancouver: NIPS, 2007: 1257-1264.
[4]
HAN P, LI Z, LIU Y, et al. Contextualized point-of-interest recommendation[C]// International Joint Conferences on Artificial Intelligence. Yokohama: IJCAI, 2020: 2484-2490.
[5]
CHANG L, CHEN W, HUANG J, et al Exploiting multi-attention network with contextual influence for point-of-interest recommendation[J]. Applied Intelligence, 2021, 51 (4): 1904- 1917
doi: 10.1007/s10489-020-01868-0
[6]
RAHMANI H A, ALIANNEJADI M, BARATCHI M, et al A systematic analysis on the impact of contextual information on point-of-interest recommendation[J]. ACM Transactions on Information Systems (TOIS), 2022, 40 (4): 1- 35
[7]
GUO L, WEN Y, LIU F Location perspective-based neighborhood-aware POI recommendation in location-based social networks[J]. Soft Computing, 2019, 23 (22): 11935- 11945
doi: 10.1007/s00500-018-03748-9
[8]
WANG H, SHEN H, OUYANG W, et al. Exploiting POI-specific geographical influence for point-of-interest recommendation [C]// International Joint Conferences on Artificial Intelligence. Stockholm: IJCAI, 2018: 3877-3883.
[9]
LI X, CONG G, LI X L, et al. Rank-geofm: a ranking based geographical factorization method for point of interest recommendation[C]// Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. Santiago: ACM, 2015: 433-442.
[10]
JIAO X, XIAO Y, ZHENG W, et al A novel next new point-of-interest recommendation system based on simulated user travel decision-making process[J]. Future Generation Computer Systems, 2019, 100: 982- 993
doi: 10.1016/j.future.2019.05.065
[11]
MANOTUMRUKSA J, MACDONALD C, OUNIS I. A contextual attention recurrent architecture for context- aware venue recommendation [C]// The 41st International ACM SIGIR Conference on Research and Development in Information Retrieval. Ann Arbor : ACM, 2018: 555-564.
[12]
ZHANG Z, LIU Y, ZHANG Z, et al Fused matrix factorization with multi-tag, social and geographical influences for POI recommendation[J]. World Wide Web, 2019, 22 (3): 1135- 1150
doi: 10.1007/s11280-018-0579-9
[13]
ZHANG J D, CHOW C Y. Geosoca: exploiting geographical, social and categorical correlations for point-of-interest recommendations[C]// Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. Santiago: ACM, 2015: 443-452.
[14]
CAI Z, YUAN G, QIAO S, et al FG-CF: friends-aware graph collaborative filtering for POI recommendation[J]. Neurocomputing, 2022, 488: 107- 119
doi: 10.1016/j.neucom.2022.02.070
[15]
JOHNSON C C Logistic matrix factorization for implicit feedback data[J]. Advances in Neural Information Processing Systems, 2014, 27 (78): 1- 9
[16]
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, 2010: 811-820.
[17]
SMIRNOVA E, VASILE F. Contextual sequence modeling for recommendation with recurrent neural networks [C]// Proceedings of the 2nd Workshop on Deep Learning for Recommender Systems. Como: ACM, 2017: 2-9.
[18]
WANG X, LIU Y, ZHOU X, et al A point-of-interest recommendation method exploiting sequential, category and geographical influence[J]. ISPRS International Journal of Geo-Information, 2022, 11 (2): 80
doi: 10.3390/ijgi11020080
[19]
ZHU H, ZHAO P, LI Z, et al. Exploiting implicit social relationship for point-of-interest recommendation[C]// Asia-Pacific Web (APWeb) and Web-Age Information Management (WAIM) Joint International Conference on Web and Big Data. Macau: Springer, 2018: 280-297.
[20]
RENDLE S. Item recommendation from implicit feedback [M]// Recommender systems handbook. New York: Springer, 2022: 143-171.
[21]
WANG H, LI L, CHE L, et al Geospatial least squares support vector regression fused with spatial weight matrix[J]. ISPRS International Journal of Geo-Information, 2021, 10 (11): 714
doi: 10.3390/ijgi10110714
[22]
CHENG C, YANG H, KING I, et al. Fused matrix factorization with geographical and social influence in location-based social networks [C]// Proceedings of the AAAI Conference on Artificial Intelligence. Toronto: AAAI, 2012: 17-23.
[23]
DACREMA M F, CREMONESI P, JANNACH D. Are we really making much progress? A worrying analysis of recent neural recommendation approaches[C]// Proceedings of the 13th ACM Conference on Recommender Systems. Copenhagen: ACM, 2019: 101-109.
[24]
STEPAN T, MORAWSKI J M, DICK S, et al Incorporating spatial, temporal, and social context in recommendations for location-based social networks[J]. IEEE Transactions on Computational Social Systems, 2016, 3 (4): 164- 175
doi: 10.1109/TCSS.2016.2631473
[25]
ZHANG J D, CHOW C Y. iGSLR: personalized geo-social location recommendation: a kernel density estimation approach [C]// Proceedings of the 21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems. Orlando: ACM, 2013: 334-343.
[26]
HE X, LIAO L, ZHANG H, et al. Neural collaborative filtering[C]// Proceedings of the 26th International Conference on World Wide Web. Switzerland: ACM, 2017: 173-182.
[27]
WANG H, TERROVITIS M, MAMOULIS N. Location recommendation in location-based social networks using user check-in data [C]// Proceedings of the 21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems. Orlando: ACM, 2013: 374-383.
