Application of Artificial Neural Network To Prediction Of Customer Order Amount In A Manufacturing System With Stochastic Process And Make-To-Order Strategy

Abstract

In this study a predictive model using Artificial Neural Network architecture was developed to predict customer orders of a manufacturing system in Nigeria with stochastic process and make-to-order strategy. This arose as manufacturing outfit in Nigeria continually contend with the problem of customer churn as a result of long waiting time of customer’s order through inability to predict the order quantity of customers on a timely basis. This necessitates the application of an Artificial Neural Network Technique on the manufacturing outfit database to predict customer’s order. A multi-layer perceptron model was built with python programming language and compared with Neuro Solution Infinity Software with close to 80 thousand dataset extracted from two databases of the manufacturing outfit in Nigeria. The results showed that the Artificial Neural Network model from Python had close performance with that obtained from the Neuro Solution Infinity software. The Root-Mean-Square Error, Mean-Absolute-Error and R2 statistical metrics were used to evaluate the performance of the developed predictive model and the results of the three metrics are 226,177 and 96.53% for the python program, and 196,123 and 98.4% for the Neuro infinity software respectively. This shows that Artificial Neural Network technique made the prediction of customer’s order amount for the product-based manufacturing outfit with stochastic process and make-to-order strategy possible.