Multi-level Population Synthesis Using Entropy Maximization-Based Simultaneous List Balancing
Synthetic population generation is the first step to an Activity Based Model (ABM). Most population synthesizers are limited when it comes to multi-level data, modifying an existing synthetic population and avoiding algorithmic errors. Monte Carlo variance resulting from drawing discrete households and persons from a probability distribution is a common source of error. In algorithms where zones are processed sequentially, errors can propagate through the list of zones resulting in large errors for the last zone processed. To the extent that these errors are higher for smaller population segments, they can adversely impact the accuracy of forecasts dependent upon these markets; for example, transit ridership estimates may be inaccurate due to errors in the location of university students. This paper presents an entropy maximization based population synthesizer (PopulationSim) which handles multiple geographies and avoids algorithmic errors. It is implemented as part of Oregon Department of Transportation’s (ODOT) effort to develop an open source population synthesis platform. PopulationSim has been implemented in the Python-based ActivitySim framework, an open-source collaborative framework for model development. PopulationSim uses a simultaneous list balancer and a Linear Programming based simultaneous integerizer to eliminate error due to the sequential processing of zones. A working version of PopulationSim was implemented for a test case and compared to a widely-used population synthesizer. PopulationSim eliminates the errors due to sequential processing of zones and results in a reasonable match to controls. Besides these major algorithmic enhancements, PopulationSim includes provision to specify flexible number of geographies and options to modify an existing synthetic population.
-
Supplemental Notes:
- This paper was sponsored by TRB committee ADB40 Standing Committee on Transportation Demand Forecasting. Alternate title: Multilevel Population Synthesis Using Entropy Maximization-Based Simultaneous List Balancing.
-
Authors:
- Paul, Binny Mathew
- Doyle, Jeff
- Stabler, Ben
- Freedman, Joel
- Bettinardi, Alex
-
Conference:
- Transportation Research Board 97th Annual Meeting
- Location: Washington DC, United States
- Date: 2018-1-7 to 2018-1-11
- Date: 2018
Language
- English
Media Info
- Media Type: Digital/other
- Features: Figures; References; Tables;
- Pagination: 18p
Subject/Index Terms
- TRT Terms: Linear programming; Population; Simulation
- Identifier Terms: Oregon Department of Transportation
- Uncontrolled Terms: Activity based modeling
- Subject Areas: Planning and Forecasting; Transportation (General);
Filing Info
- Accession Number: 01663017
- Record Type: Publication
- Report/Paper Numbers: 18-03886
- Files: TRIS, TRB, ATRI
- Created Date: Mar 20 2018 5:04PM