start with storing previous iteration demand so we can see if we converge

revise master batch file to loop until converged

re-dimension databanks
remove duplication in banks
use temporary matrices for mode choice and distribution calculations

Currently all assignments close to a best relative gap of 0.01%. This adds significant running time to the model. We want to test various options and compare run and results. Two options are:

Iterations 1,2: Best Relative Gap = 0.05%
Remaining Iterations: Best Relative Gap = 0.01%

Iterations 1,2: Best Relative Gap = 0.03%
Remaining Iterations: Best Relative Gap = 0.01%

re-dimension databanks
remove duplication in banks
use temporary matrices for mode choice and distribution calculations

1. code should not run trip generation
2. remove taz data inputs and replace with trip ends

Ideally we'd point external users of 4K to this Github repository. We need to make sure the code is scrubbed of any sensitive employment data or other items that shouldn't be in a public release.

current 4k code is not balancing school trip to attractions and is not using the balancing to school district as was done previously.

put inputs in a standard location on server

flags could include:

1. saving matrices for running bca
2. start with survey trip tables or 0 demand

Total AM boardings are 211k, counts show 124k. Need to look at Transit time of day factors, overall transit mode shares match survey fairly well.

-- Start by adjusting trip lengths to better match regional vmt
-- Look into mode choice estimation now that model skims are stabilized to lower transit share
-- Adjust trip rates as the last resort

Remove values that are extreme to avoid divide by zero errors (which Emme writes but does not stop the model run). Use same cap as SoundCast.

In the person trip table, we were having trouble comparing all the datasets. For 4k, we need to find total person trips without trucks, externals, and supplementals. Also from 4k, we'd like to know the number of external and supplemental trips. We can get that by running 4k without externals/supplementals/trucks and subtracting from the overall total.

All open source code needs a license and copyright attribution. We should:

1. Decide on a license. I suggest Apache
2. Grab generic copyright boilerplate and assign all copyright to PSRC.
3. Paste that copyright notice into every single source file.

Module 3.21 (Matrix calculations) has added a feature to the list of options so exit is no longer "4"

To avoid in the future, always exit with "q"

Trip Generation is now a stand alone "pre-process" to the 4k model. Revise code to not call trip gen bank and remove the versions of macros that call these files.

Once we figure out how we're delivering 4k model documentation, it should go in a /doc folder in this repo.