# How to calculate the approximation ratio of QAOA?

In order to evaluate the QAOA circuit, we need to compute the approximation ratio, which is the expectation value of QAOA circuit divided by the best solution.

My question is, how to find the best solution? Should we use a purely classical approach to find it (if so, which function?) or we use a hybrid classical-quantum approach with the help of a classical solver? I found in Qiskit tutorial, it uses NumPyMinimumEigensolver to obtain the result. Should we just use this result and compute its expectation value as the best solution to obtain the approximation ratio?