To my understanding algorithms like GA and PSO do not guarantee a global optimum solution and they can get stuck on a local optimum. They also have randomness in the process, so they can end up in different local optimums between runs. They may also not be very good at finding the global optimum from a plateau of reasonably good solutions.

If your problem can be expressed as a linear problem, then linear programming (LP) solvers can guarantee a global optimum and they can be fast at finding that. Other choices include mixed-integer linear programming (MILP), which can be forced to find the global optimal (but can be slow at that) and quadratic programming (QP).

Dear Shashikant Verma,

As you mentioned in your question, even if you give the same inputs to a GA algorithm, you will obtain different results each time you run the problem. This is due to the metaheuristic nature of the GA algorithms, which is exactly the reason that the results of GA algorithms are not very trustworthy compared with mathematical optimization techniques. This is simply because, on every run, each individual in the population can (and will) perform a different trajectory across the search space. However, the advantage of metaheuristic optimization over deterministic optimization is that it can solve problems that cannot be solved or need lots of time to be solved by mathematical optimization. So, in your problem, if you get different solutions after each run, you can execute your optimization several times, and pick the best solution based on the fitness values. Moreover, to get better results, you can also play with mutation, or increase the number of iterations.

Thank you, sir,

I got your point to execute optimization several times, and pick the best solution based on the fitness values in that case I can select the optimum solutions where fitness value is minimum because I performing a minimization problem

There is any possibility, after fixing all the parameters algorithm does not get solutions. Saeed Sayadi

I got your point to execute optimization several times, and pick the best solution based on the fitness values in that case I can select the optimum solutions where fitness value is minimum because I performing a minimization problem

There is any possibility, after fixing all the parameters algorithm does not get same solutions Saeed Sayadi

GA, PSO, ... and the likes should give you the same results for same input every time you run them if you control the random seed used in the process. Such seed control should be of easy access in good quality implementations.