Calibration: tour models (tour type clustering and sequence, arrival time, intermediate stops)

[ColSmith76]

Add table of tour models calibration targets and status

[ColSmith76]

Intermediate stop locations: check for any odd clustering of intermediate stops, from maps appears to be clusters north end of downtown and in part of loop -- are there TAZs that are attracting excessive numbers of stops? Should be relatively evenly distributed amongst TAZs with employment and retail employment, but never extreme.

[ColSmith76]

Left to do:

• review comparison between SEMCOG CVS and GPS tour types
• any adjustments to tour clustering and tour sequencing based on comparisons with GPS data
• arrival time at first stop, final adjustment to match GPS data
• intermediate stop location validation (See comment above regarding odd grouping)

[ColSmith76]

Left to do

• add summaries for clustering and sequencing to dashboard
• arrival time needs a bit more differentiation between vehicle types to hit targets by vehicle class (likely just a couple more iterations in calibration/tighten convergence criteria)
• add visual check of reasonableness of intermediate stop locations to dashboard

[ColSmith76]

Still an apparent cluster of "other" activity stops west side of loop and some other locations, although those look more reasonable (O hare, Woodfield), check what is driving that

[ColSmith76]

At this point the tour models are reasonably close to calibrated but there are a couple of last adjustments to make to better match the CMAP GPS data:

• more single stop tours
• more 2 stops tours and fewer > 2 stops tours amongst the multistop tours
• fewer tour clusters that only include very close stops (e.g. in same TAZ)
• currently tours are a bit too long in terms of stops and distance (too few < 30 miles, too many 30-180 miles)
• base to stop distance is okay though, so problem is the grouping of stops into tours which is creating too few tours that are too long
• please to intervene is (1) count models in stop generation -- reduce number of instances of multiple stops in same TAZ by same business and compensate by generating more stops in different TAZs to retain overall same number of stops
• allow for more single stop and low number of stop branches in the cluster algorithm

[ColSmith76]

Tours are reasonable now, except arrival time a little light on early tours (check calibration algorithm in case not matching up targets and model outputs correctly.

Check on this comment from earlier: Still an apparent cluster of "other" activity stops west side of loop and some other locations, although those look more reasonable (O hare, Woodfield), check what is driving that.

[ColSmith76]

Create some separate new issues for the two items above, not sure if they are a problem or not but do some checking in parallel with final review and sensitivity testing