Preference-Based Data Envelopment Analysis for Supplier Selection using Neural Networks

In the supply process, supplier selection and evaluation is of great importance. Among different methods to deal with this problem, Data Envelopment Analysis (DEA) is widely addressed asa successful method for solving supplier selection problem. The main advantages of DEA are to cope suiatably with qualitative dataand to perform relative camparison. However, the lack of preference toward criteria is a major drawback associated withDEA. In this paper, a new approach is introduced to consider attitude of Decision Maker (DM) called Preference Function (PF)in DEA. For each criterion, DM is requested for his preference in afinite number of points called Preference Points (PPs). A neural network (NN) is then used to approximate the PF for eachcriterion. This trained NN plays the role of PF for DEA implementation • The proposed approach of performing DEA iseasily perceptible by DMs and practitioners while it has considerable capacity for incorporating DM attitude. Finally, an experiment is conducted to compare the proposed approach with plainDEA.

Keywords:

component , Supplier Selection , Decision Maker Preference , Data Envelopment Analysis , Neural Networks