Empirical study of an artificial neural network for a manufacturing production operation

Abstract

This paper presents an empirical study of an industrial cable manufacturer in Korea. This manufacturer has also consistently been experiencing issues regarding inventory management, which have been related to production duration and the dormancy of the stock and materials. This causes unavoidable obstacles during operations, which the manufacturer cannot afford. The production orders in the case had each data set of 21 indexes, meaning a total of 21 indexes * 1,106 order samples (23,226) altogether. Two multilayer perceptron artificial neural network (MLP ANN) models were developed for the analysis. The results from two MLP ANN models successfully presented estimations for the predictive variables, these being production days (R^2 value of 0.919) and the latency days of completed products (0.773). The hierarchy of resource importance for each model was also demonstrated, which finally aims to support the judgments of small and medium-sized enterprises in regard to the inventory management. The relevance of the presented research lies in its contribution of empirical data analysis. The high number of samples contributed to making a reliable demonstration of an ANN in a practical operation system. As newly created knowledge, the data-driven advice will support the practitioners in planning inventory management, primarily when they aim to reduce the dormancy of the stock and materials by SMEs’ limited storage.