Related question about Sparse Modeling about my earlier post

[Santha]

Rob

This is a related question to my earlier post. Since its a tangential question, I am posting it as a new thread. The link to the actual thread is here

Just a tangential question of the same model: This is about how sparse modeling. In the sample set I gave you for my actual problem, the cost from IC to DC is different. that can be one flavor of the model and is totally fine to have that. 

But, there is another flavor where the cost from IC to dest_code is always the same. See table below. It is $100 per kg from BUD_A to JFK.  Just that there are different Dest_Codes and so I have it as $100 for each possible combination of it. While this is not wrong, can we make this more efficient by building a sparse set of {BUD_A to JFK}?  If yes, how do it do it? The first thought that came to my mind was if the cost is $100 for all the dest_code when why can't we have one dest_code and say JFK instead of JFK_N_OP_3 etc etc and be done with it. I need the granular definition of dest_code for me to compare what happened originally vs what SAS model gave. So , that is where I am coming from. Please help me with you thoughts.

Mot1	Dest_Code	Outbound_Rate
BUD_A	JFK_N_OP_3	100
BUD_A	JFK_N_SY_3	100
BUD_A	JFK_N_CH_3	100
BUD_A	JFK_N_RA_3	100
BUD_A	JFK_N_SY_2	100
BUD_A	JFK_N_OP_2	100

[RobPratt · Accepted solution]

If various dest_codes can be merged into one when they have the same costs, that sounds like a good way to reduce the model size.  You can do this in the input data creation steps rather than changing the optimization model.  Any comparisons between solutions with different granularity can be done in postprocessing, outside of PROC OPTMODEL.

[Santha]

Thanks Rob. I will do that if we are going to do that flavor of the model where the various dest_codes can be merged. thank you.