K-factors

[isaac-tfn]

Add in the option to better match distributions to observed post-gravity model. This could potentially be in the form of a triply constrained furness on the final run (i.e. when cost-function factors have been determined, on the final furness constrain to match distributions as well as to match target rows and columns).

[nhannguyen2212]

1. Method 1:

   • Calibrate mu/sigma as per current process in MPV -> produce T[i,j]
   • Produce achieved TLD[k] from the post-calibrated matrices T[i,j] and Dist[i,j]
   • Calculate TLD adjustment factor TLD_adj[k] = TLD_target[k] / TLD_achieved[k]
   • Apply TLD adjustment to post-calibrated matrix: T1[i,j,k] = T0[i,j,k] * TLD_adj[k]
   • Furness doubly-constrained on T1[i,j] with targeted O/D trip-ends: T[i,j] = A[i] B[j] O[i] D[j] T1[i,j,k]

2. Method 2:

   • Calibrate mu/sigma as per current process in MPV -> produce T[i,j]
   • Produce achieved TLD[k] from the post-calibrated matrices T[i,j] and Dist[i,j]
   • Calculate TLD adjustment factor TLD_adj[k] = TLD_target[k] / TLD_achieved[k]
   • Apply TLD adjustment to Furness process: T[i,j] = A[i] B[j] TLD_adj[i,j,k] O[i] D[j] * f(C[i,j])