-
Notifications
You must be signed in to change notification settings - Fork 99
Project Meeting 2020.04.03
Ben Stabler edited this page Apr 6, 2020
·
31 revisions
Task 1 Project Management (milestone)
- Discussed draft Phase 6 scoping Google doc here
- Comments due in two weeks, 4/17
- Plan to begin to prioritize in second half of April
- Make sure we get this process into the strategic development plan
Task 2 Strategic Planning (milestone)
- Discuss strategic development and contribution plan outline here
- Comments due in two weeks, 4/17
Task 4 I/O Improvements (milestone)
- No updates
Task 5 TNCs (milestone)
- Summarized updated TNC example from MTC
- Updated ActivitySim expressions
- Need to run full sample and compare
- Use the verification summaries from last time
Task 6 Model Developer Tutorial (milestone)
- No updates
Task 9 Estimation Mode (milestone)
- Discussed progress to date:
- New estimation.yaml config file and estimation example being built
- Each sub model gets an override choice inputs table with either HHs, persons, tour, or trips and a choice column override
- Lots of work done to clean-up expressions so coefficients are separate from data
- Working with sanitized data so we don't use resources cleaning specific survey data
- Software is quite agnostic with respect to data fields so functionality is flexible
- User will input a household, person, tour, trip table and activitysim can also output the same thing (it currently outputs this + additional intermediate data fields as well)
- Estimation data bundle (EDB) built for auto ownership, work location, tour mode choice
- EDB includes coefficient file, model specification file, data values combined file (chooser data and alternatives data), and choices file
- Working through all submodel now and restructuring inputs + adding estimation integration code
- Python estimation tools reviewed and going forward with larch
- Next steps
- Jeff continue to work though code, Ben to work on estimation integration scripts
- About 2/3 of task budget spent
- How much bad data, error checking will be included? Do we toss bad records? For now plan is to have more articulate logging around input data issues.