Branching

Branching Priorities (aka Priority Orders)

Solvers enable programmers to set preferable branching orders on each Variable. By this order, the solver selects the next variable to branch on. This might drastically reduce solution times. When setting variable branching orders, keep in mind

give higher priority to variables

• that do not depend on other branchable (Integer or Binary) variables
• that are earlier in time (if any kind on time / time period concept is in the model)

Note: Higher values for Variable.BranchingPriority indicate that those variables are selected with preference over variables with a lower value.

• Cf. the respective documentation of
  • CPLEX Priorities
  • Gurobi BranchPriority

Branching Directions

The BranchingDirection of a Variable gives a preferred branching direction for the solver. In case of a fractional integer or binary Variable, the solver usually chooses based on the fraction (and some additional details) whether to branch up or down. With a non-default branching direction, this automatic choice is overridden.

The direction is of Default, Up or Down.

Note: For an overview which Solver supports Branching Priorities, refer to the supported solvers page

Branching details

Find background information about MIP Tuning and Branching in http://www-01.ibm.com/support/docview.wss?uid=swg21400023