Based on historical bike counting information, geographical and temporal patterns in human mobility can be detected. Predicting bicycle traffic and traveler flows enable the identification and prevention of potential bottlenecks in a city's cycling network and creates new opportunities for mobility solutions. Since the introduction of the first bicycle counting station in Brussels in 2017, the city has expanded its counting network to twelve stations and is aiming to reach fifteen stations by the end of 2019. Real-time and historical bike counting data concerning these stations is made available to the public through web endpoints. In this paper, we introduce BRUBIKE, a novel aggregated dataset of bicycle and meteorological information concerning the city of Brussels. We aim to lower the boundary of accessing Brussels' cycling information and to stimulate the creation and evaluation of novel traffic flow models on Brussels' data. A subset of existing machine learning models is evaluated on the proposed dataset with the task of predicting bicycle traffic for a yet unseen period, once with weather parameters, and once without weather parameters. Results indicate significantly better prediction performance when weather parameters are included due to the existing correlation of weather and bike traffic. Finally, we propose an open source application to make historical bike traffic and predictions more accessible towards Brussels' citizens.
Vanden Broucke, S, Vidal Piña, LM, Do Huu, T & Deligiannis, N 2019, BRUBIKE: A dataset of bicycle traffic and weather conditions for predicting cycling flow. in IEEE International Smart Cities Conference. pp. 432-437, the IEEE International Smart Cities Conference, Casablanca, Morocco, 14/10/19.
Vanden Broucke, S., Vidal Piña, L. M., Do Huu, T., & Deligiannis, N. (2019). BRUBIKE: A dataset of bicycle traffic and weather conditions for predicting cycling flow. In IEEE International Smart Cities Conference (pp. 432-437)
@inproceedings{8c135b4d85984282ad77585951c7601d,
title = "BRUBIKE: A dataset of bicycle traffic and weather conditions for predicting cycling flow",
abstract = "Based on historical bike counting information, geographical and temporal patterns in human mobility can be detected. Predicting bicycle traffic and traveler flows enable the identification and prevention of potential bottlenecks in a city's cycling network and creates new opportunities for mobility solutions. Since the introduction of the first bicycle counting station in Brussels in 2017, the city has expanded its counting network to twelve stations and is aiming to reach fifteen stations by the end of 2019. Real-time and historical bike counting data concerning these stations is made available to the public through web endpoints. In this paper, we introduce BRUBIKE, a novel aggregated dataset of bicycle and meteorological information concerning the city of Brussels. We aim to lower the boundary of accessing Brussels' cycling information and to stimulate the creation and evaluation of novel traffic flow models on Brussels' data. A subset of existing machine learning models is evaluated on the proposed dataset with the task of predicting bicycle traffic for a yet unseen period, once with weather parameters, and once without weather parameters. Results indicate significantly better prediction performance when weather parameters are included due to the existing correlation of weather and bike traffic. Finally, we propose an open source application to make historical bike traffic and predictions more accessible towards Brussels' citizens.",
keywords = "Data visualization, Big Data applications, machine learning",
author = "{Vanden Broucke}, Steven and {Vidal Pi{\~n}a}, {Luis Manuel} and {Do Huu}, Tien and Nikolaos Deligiannis",
year = "2019",
language = "English",
pages = "432--437",
booktitle = "IEEE International Smart Cities Conference",
note = "the IEEE International Smart Cities Conference : (ISC2 2019) ; Conference date: 14-10-2019 Through 17-10-2019",
}