Originally posted by: Mortsnets

We have a maximization problem that refuses to solve. The problem file is too large to attach.

As soon as it switches from scaled infeasible to objective iterations it has a huge objective function:

Iteration: 38105    Scaled infeas =             0.000015
Iteration: 38233    Objective     = 7223476683799698300000.000000
Elapsed time =   18.75 sec. (38233 iterations)
Iteration: 38683    Objective     = 7369411016531076200000.000000
Iteration: 38686    Scaled infeas =       3413594.075667
Iteration: 39276    Scaled infeas =          5458.253910
Switched to steepest-edge.
Iteration: 39847    Scaled infeas =            55.782989
Removing shift (7969).
Markowitz threshold set to 0.5

Here are the stats:

Variables            :  195760  [Nneg: 56664,  Fix: 10496,  Box: 380, Free: 128100,  Other: 120]
Objective nonzeros   :      12
Linear constraints   :  179225  [Less: 8148,  Greater: 11820,  Equal: 158844, Range: 413]
  Nonzeros           : 2406313
  RHS nonzeros       :     329

Variables            : Min LB: -1.000000e+017   Max UB: 1.000000e+017
Objective nonzeros   : Min   : 1.000000         Max   : 1.000000
Linear constraints   :
  Nonzeros           : Min   : 4.336809e-019    Max   : 147254.8
  RHS nonzeros       : Min   : 0.07250000       Max   : 1.000000e+020
  Range values       : Min   : 0.0000000        Max   : 1.000000e+017

Any ideas what is causing it to struggle so much?

Thanks,

Craig

What exactly do you mean by "refuses to solve"? Is CPLEX stuck? Does it crash?

The statistics for your model do not look good. Your coefficients range from 1e-19 through 1e6 and the range of other data in your problem is huge as well. This is certainly a cause of trouble for any solver using floating point arithmetic. Any chance you can narrow down those ranges significantly?