Development of an ensemble learning algorithm based on neuralnetwork multi-objective optimization in supplier selection: a casestudy, SAPCO Parts Supply Company

Nowadays, one of the key components in the supply chain is the evaluation and selection ofsupply chain management. Taking into account many effective factors in monitoring and decisionmakinghas turned this problem into a multi-objective problem, on the other hand, the existence ofcompetition makes finding suppliers of materials, goods and semi-finished parts difficult. Because,with the expansion of societies, there are more options to choose business partners. It is clear thattaking this managerial decision with traditional statistical and mathematical methods, will not beefficient.Studies show that modern technologies and innovations, including methods based on artificialintelligence and machine learning, which have the ability to solve complex models, have performedbetter. In this regard, in this research, a multi-objective optimization algorithm MOEA/D is usedto find the optimal architectures as the input of the neural network. Next, after obtaining the Paretofront including the most optimal architectures, each one as a neural network different from theother forms the basic models in a collective solution algorithm. It is clear that the simultaneous useof several decision-making networks can significantly increase the prediction accuracy.The performance of the proposed model was evaluated based on different evaluation criteriaon a dataset from IKCO Engineering Design and Parts Supply Company, SAPCO, which is thelargest industrial group in Iran. The obtained results indicate the high accuracy of the proposedmodel compared to other methods for predicting the efficiency of each of the suppliers in thiscompany.

Keywords:

Supply chain management , data envelopment analysis , artificial neural networks , ensemble learning