CPLEX returns optimal solution for small instances of a MIP model. But when try to solve a larger instance, CPLEX return a local optimum (solution checked by a heuristic). The model is MIP with a number of Big-M formulations that are replaced by Indicator Constraints. For one of the large instances:
# of constraints: 2086337
#of variables: 1122331 (66225 int)
#of none zero: 7223176
It takes 53557 seconds for CPLEX to close the optimality Gap.
Any idea? please share your thought.