Originally posted by: walterOR

I just noticed that in the Python API for Gurobi there is a method that can be used to set the branching priority for each variable. Not completely sure to me how it work in detail, but it seems to be a useful feature for certain problems.

Is there anything comparable in Cplex (docplex for Python, preferably)?

Hi,

in python you could have a look at the example mipex3.py which illustrates SOS sets and priority orders.

in OPL if I use the example from https://www.linkedin.com/pulse/what-optimization-how-can-help-you-do-more-less-zoo-buses-fleischer/

I can write

int nbKids=300;
float costBus40=500;
float costBus30=400;
 
dvar int+ nbBus40;
dvar int+ nbBus30;

execute
 {
 nbBus40.priority=10;
 nbBus30.priority=1;
 }
 
minimize
 costBus40*nbBus40  +nbBus30*costBus30;
 
subject to
{
 40*nbBus40+nbBus30*30>=nbKids;

}

which changes some decision variables priorities

regards

Adding to what Alex wrote, for the CPLEX Python API, you can use the order interface.

Underneath the hood, this uses CPXXcopyorder from the C Callable Library.

So with docplex , if I follow

https://www.linkedin.com/pulse/making-optimization-simple-python-alex-fleischer/

I could write

from docplex.mp.model import Model

def set_ordering(dvars, weights, brdirs):
    # set priority ordering on a collection of variables,
    # given a list of weights, and a list of -1,0,1 directions
    ldvars = list(dvars)
    lweights = list(weights)
    ldirs = list(brdirs)
    if ldvars:
        m = ldvars[0].model
        # check all lists have same length
        # weights are numbers
        # directions are -1, 0, 1

        cpx = m.get_cplex()
        cpx.order.set([(dv.index, w, brd) for dv, w, brd in zip(ldvars, lweights, ldirs)])
        cpx.order.write('%s_prio.ord' % m.name)

mdl = Model(name='buses')
nbbus40 = mdl.integer_var(name='nbBus40')
nbbus30 = mdl.integer_var(name='nbBus30')

set_ordering(dvars=[nbbus40, nbbus30], weights=[10, 1], brdirs=[0, 0])
                        
mdl.add_constraint(nbbus40*40 + nbbus30*30 >= 300, 'kids')
mdl.minimize(nbbus40*500 + nbbus30*400)

mdl.solve()

for v in mdl.iter_integer_vars():
   print(v," = ",v.solution_value)

Originally posted by: walterOR

Thank you for these suggestions. This is exactly what I was looking for.