estimating the parameters of logit model using simulated annealing algorithm: case study of mode choice modeling of isfahan

logit models are one of the most important discrete choice models and they play an important role in describing decision makers’ choices among alternatives. in this paper the multi-nominal logit models has been used in mode choice modeling of isfahan. despite the availability of different mathematical computer programs there are not so many programs available for estimating discrete choice models. most of these programs use optimization methods that may fail to optimize these models properly. even when they do converge, there is no assurance that they have found the global optimum, and it just might be a good approximation of the global minimum. in this research a heuristic optimization algorithm, simulated annealing (s.a), has been tested for estimating the parameters of a logit model for a mode choice problem that had 17 parameters for the city of isfahan and has been compared with the same model calculated using gauss that uses common and conventional algorithms. simulated annealing is and algorithm capable of finding the global optimum and also it’s less likely to fail on difficult functions because it is a very robust algorithm and by writing the computer program in matlab the estimation time has been decreased significantly. in this paper, this problem has been briefly discussed and a new approach based on the simulated annealing algorithm to solve that is discussed and also a new path for using this technique for estimating nested logit models is opened for future research by the authors. for showing the advantages of this method over other methods explained above a case study on the mode choice of isfahan has been done.