Fighting Traffic Congestion with Exploratory Data Analysis

Paul Cote
Massachusetts Institute of Technology
1992

This project represents an experiment in making tools to make sense of a very complicated database through maps. The database was the 1980 Census Transportation Planning Package for Eastern Massachusetts; a block-group level origin-destination (OD) matrix. This database has a record for each of the 1292 block-groups in the Boston Metropolitan region. For each of these blockgroups, the OD matrix has a number of trips, by choice of transportation mode, to each of the other block groups. THe information in this database came from the 1980 Census Long Form which is a sample of one-sixth of all households.

How can this massive amount of information be used to better understand the choices that commuters make across this region? TO answer this question, the OD matrix was loaded into SYBASE, an industrial-grade relational database management system, a series of summaries and queries were made, and the results transferred to ArcInfo, a geographic information system for mapping.

Regional Commuter Choices

Are there aspects of the residential environment that affect the mode choices of commuters? The vertical symbols on this map reflect the number of commuter trips and proportioal mode-split for people commuting into Suffolk County from each of the towns in Eastern Massachusetts.

Mode Choices by Commuter Destination

Are commuter mode choices affected by qualities of the destination environment?

Worker Density in Suffolk County

The Transportation Planning Package provides an uncommon glimpse of the daytime population, which in central Boston is much different than the residential population exhibited by ordinary census data. This map splits Suffolk County into four areas, each being the destination of one quarter of the total commuters coming into the county. It is clear that some destinations are very dense.

Mode Choices by Density of Commuter Destination

Does the density of workers at a commuter's destination play a role in determing the choice of transportation modes? These pie charts take the same segments of the commuter population that are reflected in the quartile categories of the previous map, and summarizes the mode choices for each group. Evidently, people who work in the very dense areas are much less likely to take their own car to work.

This project was done as a one-semester independent study in 1992. While looking at these static maps, it is easy to see how many useful questions might be asked with a more dynamic interface to the data.