msmobility / mito

Microscopic Transportation Orchestrator
7 stars 13 forks source link

How to interpret travel time budget UEC model results? #13

Closed nkuehnel closed 7 years ago

nkuehnel commented 7 years ago

In TravelTimeBudget we set the HBW budget according to travel time by car (similiar for HBE, once #8 is fixed). For adjusting all other purposes we subtract HBE and HBW budget from the total travel time budget and distribute what's left (whith a minimum of 0) :

double discretionaryTTB = totalTravelTimeBudget -
                household.getTravelTimeBudgetForPurpose(Purpose.HBW) -
                household.getTravelTimeBudgetForPurpose(Purpose.HBE);

discretionaryTTB = Math.max(discretionaryTTB, 0);

The UEC travel time budget calculator usually returns values around 3-6. The skim matrix for travel time returns the time in minutes like (60 minutes). In that case every household gets a discretionary budget of 0, as the subtraction will almost always result in a value below 0.

I guess the result from the UEC has to be converted somehow. However, just assuming it's result being in hours unit doesn't seem to be right, as the total travel time would often be around 4 or 5 hours even for a single household. How is the result of the travel time budget UEC model to be interpreted?

nkuehnel commented 7 years ago

The results of the excel sheet calculator only resembled the utilities for the underlying weibull model:

image

where everything inside exp(...) is the utility. The program itself expects the resulting travel time budget, not the utility, which should be calculated inside the UEC.