Supplier Selection Using Fuzzy Analytic Network Process Based On Artificial Neural Network

Utilizing resources in the most productive way possible and purchasing goods and services at the lowest price available are two vital topics for companies.This study proposes a supplier selection ?decision model? for a company or corporation. The main purpose of this study is to evaluate supplier alternatives with respect to identified processes and hence to determine the best supplier. For this purpose, Artificial Neural Networks (ANN) and Fuzzy Analytic Network Process (FANP) were used in this study. Although Analytic Network Process, which is commonly used for multi criteria decision making problems, takes into consideration both objective-subjective criteria and is a widely used technique, it is not sufficient in illustrating the decision maker?s judgment and accounting for and quantifying the uncertainty in the problem. Therefore, FANP was utilized in compensating for the issue of uncerainty. Chang?s extend analysis method was used for fuzzy pairwise comparison matrix solution.During the development of the supplier selection model, in the phase of forming fuzzy pairwise comparison matrices, we need supplier selection experts? or decision makers? opinions. However, if some change occurs in the model criteria, available suppliers, or the conditions that affect supplier selection, experts have to reconvene for decision making, and this causes some problems. In order to remove this problem fuzzy pairwise comparison matrix data that indicates experts opinions was used in different ANN models and also the data was used in the learning stage. Thus even if there is a change in pairwise comparison values, there is no need to re-consult the experts.One advantage of using ANN for supplier selection problems is the neural network?s learning ability. When this model is applied for a different selection problem (it is assumed that the model has same criteria), it is possible to determine the criteria weights using ANN without any need to apply Chang?s approach. In case a change made by the decision makers on pairwise comparison matrices, new weights are found in a very short time and this is another benefit that the model provides. This feature provides the decision makers the opportunity to make decisions without the need to worry about computation time in complex models.Key words: Supplier selection, Fuzzy Analytic Network Process, Artificial Neural Network