Dr. Hamdi Kavak: Spatiotemporal Prediction of Foot Traffic

Spatiotemporal Prediction of Foot Traffic

Mentors

Taylor Anderson
Hamdi Kavak
Tim Leslie
Amira Roess
Andreas Züfle

Student and Senior Collaborators

Samiul Islam (CSI Ph.D. student - 2021-current)
Dhruv Ghandi (CDS undergraduate student - STIP 2021)
Justin Elarde (CDS MS student)

Project Dates

2021-current

Initial Work

This project started as part of STIP 2021 and aims to predict future foot traffic: the number of people from each census block group (CBG) that will visit each POI of a study region with potential applications in marketing and advertising. Specifically, we explore different techniques to predict weekly foot traffic data at the POI level. We propose a collaborative filtering approach using tensor factorization, which provides a de-noised estimation of visits in previous weeks for all POI-CBG pairs. Using this tensor, we explore various time series prediction models: weekly rolling average, weighted weekly rolling average, univariate linear regression, polynomial regression, and long short-term memory (LSTM) recurrent neural networks. Our initial results show that collaborative filtering consistently improves the prediction results of all the prediction models. We also found that a simple weighted average always performed better than the more sophisticated approaches. Given this abundance of foot traffic data, this result shows that we can improve the spatiotemporal prediction of foot traffic data by harnessing collaborative filtering.

Extensions

Currently, we are improving the validation of the initial results and try to extend the study to other regions.

Publications & Presentations

Spatiotemporal Prediction of Foot Traffic
S. Islam, D. Gandhi, J. Elarde, T. Anderson, A. Roess, T.F. Leslie, H. Kavak, and A. Züfle
5th ACM SIGSPATIAL International Workshop on Location-Based Recommendations, Geosocial Networks, and Geoadvertising, Seattle, Washington, USA (Online), November 2-5, 2021
[Paper] [BibTex]