Bike sharing systems, usage mining.

The bicycle sharing systems are increasingly numerous nowadays. These transportation systems generate sizable transportation data the mining of which can reveal the underlying urban phenomenons linked to city dynamics.

Goal

BSS usage data may give insights on the relationships between stations neighborhoods type (the amenities it offers, its sociology, ...) and the observed mobility. A first method developped in this pre-print, enables the discovery of regions of different functions, that induce specific usage patterns in BSS data. These potentials are demonstrated through an in-depth analysis of the results obtained on the Vélib’ large-scale bike sharing system of Paris. We also initiate another line of work that use LDA for dynamical Origin / Destination matrices analysis. Some first results obtained with such an approach are detailed in the slides.

Insight

The first work introduces a statistical model to automatically analyze bike sharing system trips data. This model introduce a latent variable to partition the stations in terms of their temporal dynamics over the day with respect to the number of rented and returned bikes. This generative model is based on Poisson mixtures and introduces a station scaling factor that handles the discrepancy between the stations activities. Eventually, the difference of dynamics between week days and week-end is also taken into account. This model try to find the latent factors that shape the geography of trips. We also initiat another line of work that use LDA for dynamical Origin Destination matrix analysis.

Ressources:

Paper

E. Côme, L. Oukhellou. Model-based count series clustering for Bike-sharing system usage mining, a case study with the Vélib’ system of Paris. Submited to ACM TIST. [pdf-preprint]
This paper introduces a statistical model to automatically analyze bike sharing system trips data. This model will introduce a latent variable to partition the stations in terms of their temporal dynamics over the day with respect to the number of rented and returned bikes. This generative model is based on Poisson mixtures and introduces a station scaling factor that handles the discrepancy between the stations activities. Eventually, the difference of dynamics between week days and week-end will also be taken into account. This model will and the latent factors that shape the geography of trips.


Paper

E. Côme, A. Randriamanamihaga, L. Oukhellou and P. Aknin. Spatio-temporal analysis of Dynamic Origin-Destination data using Latent Dirichlet Allocation. Application to the Vélib’ Bike Sharing System of Paris. Accepted TRB 2014. [pdf-preprint]
This paper deals with a data mining approach applied on Bike Sharing System Origin-Destination data, but part of the proposed methodology can be used to analyze other modes of transport that similarly generate Dynamic Origin-Destination (OD) matrices...


Paper

A. Randriamanamihaga, E. Côme, L. Oukhellou and G. Govaert. Clustering the Vélib’ origin-destinations flows by means of Poisson mixture models. Application to the Vélib’ Bike Sharing System of Paris. Accepted Neurocomputing. [pdf-preprint]
Studies based on human mobility, including Bicycle Sharing System analysis, has expanded over the past few years. They aim to give insight of the underlying urban phenomena linked to city dynamics. This paper presents a generative count-series model using adapted Poisson mixtures to automatically analyse and find temporal-based clusters over ...


Paper

Y. Han, E. Côme, L. Oukhellou. Towards bicycle demand prediction of large-scale bicycle sharing system. [pdf-preprint]
We focus on predicting demands of bicycle usage in Velib system of Paris, which is a large-scale bicycle sharing service covering the whole Paris and its near suburbs. In this system, bicycle demand of each station usually correlates with historical Velib usage records at both spatial and temporal scale. The spatio-temporal correlation acts as an important factor affecting bicycle demands in the system. Thus it is a necessary information source for predicting bicycle demand of each station accurately. To investigate the spatio-temporal correlation pattern and integrate it into prediction...


Visual exploration of live and historical data

This application offers an access to historical data about some twenty BSS systems, such as Paris, London or New-York at different temporal and spatial scales. It was developped at ifsttar during a research project on BSS data analysis: we think that an historical vision is mandatory to better analyze, understand and eventually enhance these systems. The open-data movement has lead an increased number of city to open real time stock data (see jcdecaux open data site for examples). We therefore take this opportunity to build an historical database where these real time data are recorded. This site is a first attempt to visualize these data, it is organized around three views offering the possibility to compare the systems, to observe a city more closely or a given station.

Dataviz

This dataviz offers to explore one month of Velib trips data. You may select a time frame, and see flow patterns per stations. Stations can also be filtered by the type extracted using the clustering procedure we have working on.

Dataviz

This dataviz offers to explore one month of Velib trips data. You may move between time frame, to explore the density of generated trips over paris.

Animated Map

Just a map moving in time showing the evolution of Paris BSS Vélib' stocks over a week to see comute patterns. Be carefull in the interpretation the stocks are not only influenced by users move but also by the bike redistribution performed by JCdecaux.

Map

Since I work mainly on bike share nowadays. I wanted to make a "cyclist" map who will higlight bike equipments ("the cyclability of ways") and relief (even low). And like many things on this issue are currently done (see tender of the island region of France, arrived from base map google maps "Bike", ...), I wanted to see what could be done with open source tools, open data, or at least data I had access to and a little goodwill. The current results of my tests is available here.