Originally posted by: tpd1

I would like to make sure I understand settings for branching order:
1: decreasing cost coefficients among the variables

For setting 2, CPLEX gives higher branching priority to variables with smaller bound range for example 0 <= x1 <= l1,  0<= x2 <= l2, l1<=l2 so x1 get higher priority for branching ?

I am not sure about setting 3, could you guys point me to some reference ? 

Originally posted by: tpd1

I found more explanations for the settings here:

http://ampl.com/BOOKLETS/amplcplex90userguide.pdf

Originally posted by: PierreBonami

Hi,

The description in the AMPL guide is accurate.

For setting 1, CPLEX gives higher priority to variables with higher cost (not reduced cost).

For setting 2, CPLEX gives higher priority to variables with smaller range

Setting 3, gives higher priority to variables where cost divided by the number of non-zero matrix coefficients is smaller (the ampl guide gives a good intuition of why this might be a good idea).

Note however that setting this parameter is optional and by default CPLEX relies on the methods set by the following parameter to choose the branching variable:
http://www.ibm.com/support/knowledgecenter/en/SSSA5P_12.6.3/ilog.odms.cplex.help/CPLEX/UsrMan/topics/discr_optim/mip/performance/21_var_seln.html

I hope this helps.

Pierre