01660385

Component

The authors present a semi-supervised learning model to estimate the daily professional and scholar mobility flows from static censuses. Speedy urbanization has caused the estimation of the human mobility flows an essential step in pursuit of transport and urban planning. The principal purpose of this study is to transform professional and scholar stationary censuses data into a dynamic origin destination matrix form. This provides the opportunity to merge them with other mobility data. The professional and scholar mobility flows are among the most weekly regular displacements culminating into road congestion problems. Understanding their dynamic evolution along the day plays a pivoted role in assisting decision making processes for local authorities. Professional and scholar censuses often comprised home and work/educational institution information, albeit temporally aggregated. The authors thus propose a neural network model that learns the temporal distribution of displacements across other mobility sources attempting to predict them on new censuses data. The authors' approach has been validated in a real context across 8,000 features with three prime advantages. First the model can be built from a large amount of widely available unlabeled census data with the prerequisite of limited labeled inputs (as old surveys). The latter are used only to fine-tune the model for inferring the origin destination matrices. Second, the resulting matrices inherit all the trip information and the socio-demographic attributes of the individuals contained in the aggregated census. Finally, the model has the scalability across other French towns even if no old survey upon it is available.

This paper was sponsored by TRB committee ADB40 Standing Committee on Transportation Demand Forecasting.

18-02805

English

15p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; References; Tables

Neural networks; Origin and destination; Traffic counts

Restricted Boltzmann Machine

France

Highways; Operations and Traffic Management; Planning and Forecasting

Transportation Research Board Annual Meeting 2018 Paper #18-02805

TRIS, TRB, ATRI

Jan 8 2018 10:40AM