Contexts on POI

Findings

Focus on Contexts, Metrics, and Models

Focus on Contexts

Considering all contextual information in a model will not necessarily improve the performance.

Focus on Metrics

The performance of models varies when compared based on different evaluation metrics.

Focus on Models

The geographical or temporal information is more effective in improving the M model.

Sources

Yelp

Baselines	Contexts	Pre@5	Pre@15	Rec@5	Rec@15	nDCG@5	nDCG@15
GeoSoCa	G	0.0112	0.0105	0.0085	0.0245	0.0109	0.0106
	S	0.0135	0.0120	0.0084	0.0214	0.0136	0.0125
	C	0.0018	0.0017	0.0013	0.0037	0.0018	0.0018
	GS	0.0193	0.0152	0.0111	0.0262	0.0194	0.0165
	GC	0.0165	0.0149	0.0120	0.0332	0.0166	0.0155
	SC	0.0135	0.0119	0.0082	0.0218	0.0136	0.0124
	GSC	0.0211	0.0162	0.0120	0.0277	0.0215	0.0179
FCFKDEAMC	G	0.0109	0.0104	0.0082	0.0235	0.0109	0.0106
	S	0.0180	0.0141	0.0101	0.0231	0.0181	0.0153
	T	0.0165	0.0152	0.0119	0.0324	0.0172	0.0159
	SG	0.0198	0.0157	0.0111	0.0266	0.0199	0.0169
	ST	0.0194	0.0158	0.0116	0.0261	0.0201	0.0172
	GT	0.0208	0.0183	0.0150	0.0391	0.0210	0.0192
	SGT	0.0208	0.0168	0.0122	0.0277	0.0215	0.0183
PFMMGM	M	0.0085	0.0080	0.0045	0.0141	0.0084	0.0081
	G	0.0076	0.0074	0.0057	0.0165	0.0074	0.0073
	MG	0.0188	0.0143	0.0144	0.0325	0.0193	0.0158

Gowalla

The gowalla dataset does not include the cateogorical information.
Baselines	Contexts	Pre@5	Pre@15	Rec@5	Rec@15	nDCG@5	nDCG@15
GeoSoCa	G	0.0156	0.0149	0.0095	0.0270	0.0155	0.0150
	S	0.0217	0.0192	0.0117	0.0288	0.0221	0.0202
	GS	0.0230	0.0203	0.0138	0.0355	0.0233	0.0213
FCFKDEAMC	G	0.0164	0.0160	0.0098	0.0285	0.0162	0.0160
	S	0.0345	0.0263	0.0170	0.0369	0.0353	0.0292
	T	0.0581	0.0455	0.0283	0.0640	0.0593	0.0499
	SG	0.0385	0.0298	0.0192	0.0417	0.0391	0.0327
	ST	0.0573	0.0404	0.0273	0.0526	0.0596	0.0465
	GT	0.0559	0.0437	0.0275	0.0620	0.0568	0.0478
	SGT	0.0458	0.0391	0.0260	0.0507	0.0571	0.0450
PFMMGM	M	0.0191	0.0147	0.0084	0.0194	0.0231	0.0182
	G	0.0196	0.0176	0.0111	0.0291	0.0198	0.0184
	MG	0.0275	0.0222	0.0149	0.0356	0.0281	0.0242

Extra

Cite

@article{rahmani2021contextsPOI,
  author =  {Hossein A. Rahmani, Mohammad Aliannejadi, Mitra Baratchi, Fabio Crestani},
  title = {A Systematic Analysis on the Impact of Contextual Information on Point-of-Interest Recommendation},
  journal = {ACM Transactions on Information Systems (TOIS)},
  volume = {-},
  year = {2022},
  url = {https://arxiv.org/pdf/2201.08150.pdf},
  archivePrefix = {arXiv},
  eprint =  {XXX}
}

About Us

Our Group

Hossein A. Rahmani is a visiting researcher at the WI group, Universtiy College London (UCL). Previously he was a visiting research at USI-IR, Universita della Svizzera Italiana (USI) during this project. He recived his BSc in Software Engineering at the University of Zanjan, Iran, in 2016. Currently, he is a master's student at the University of Zanjan, Iran. He is interested in problems at the intersection of Machine Learning, Data Analysis and BigData as well as Data Mining. He is currently working on Point-of-Interest recommender systems, a project to recommend locations in location-based social networks.

Hossein A. Rahmani - Universtiy College London (UCL)

h.rahmani@ucl.ac.uk

Mohammad Aliannejadi is a post-doctoral researcher at IRLab (formerly known as ILPS), the University of Amsterdam in The Netherlands. He obtained his Ph.D. in the Faculty of Informatics, Università della Svizzera italiana (USI) in Lugano, Switzerland. During his Ph.D., he spent four months visiting CIIR at the University of Massachusetts, Amherst, USA. Before his Ph.D., he completed his MSc degree in the Department of Computer Engineering and Information Technologies at Tehran Polytechnic, Tehran, Iran.

Mohammad Aliannejadi - University of Amsterdam

m.aliannejadi@uva.nl

Mitra Baratchi works as an assistant professor in the Leiden Institute of Advanced Computer Science (LIACS) at Leiden University. She is the founder of the Special Interest Group on Spatio-Temporal Data Mining (SIG-SDTM) and member of the Automated Design of Algorithms (ADA) research group. Previously, she worked as a post doctoral researcher in the Design and Analysis of Communication Systems (DACS) group at University of Twente in the Living Smart campus project.

Mitra Baratchi - Leiden University

m.baratchi@liacs.leidenuniv.nl

Fabio Crestani is a full professor at the Faculty of Informatics of USI since January 2007 and head of USI-IR research group on information retrieval. Before arriving in Lugano he was a (full) Professor at the University of Strathclyde in Glasgow (UK) since 2000. During that time he was a Visiting Professor at IMAG (France), and spent a year sabbatical at UC Berkeley (USA) and Xerox PARC (USA). In 1997-1999 he was a Postdoctoral Research Fellow at the University of Glasgow (UK), at the International Computer Science Institute in Berkeley (USA), and at the Rutherford Appleton Laboratory (UK).

Fabio Crestani - Universita della Svizzera Italiana

fabio.crestani@usi.ch

Contextual Information