Comparison of Artificial Neural Network and Support Vector Machine on Production Quantity Prediction

Abstract

This research presents the comparison of Neural Network Back Propagation (NNBP) model and Support Vector Machine (SVM) for predicting production quantity. This model is built based on input variables that affect the determination of production quantity which include demand, setup costs, production, material costs, holding costs, transportation costs. The performance of NNBP and SVM can be anlyzed using Root Mean Square Error (RMSE). The experiment is performed by optimizing the parameter of NNBP model and SVM by trial and error to find the smallest error between actual and predicted. The proposed models are examined using primary dataset that was collected from Iron Casting Manufacturing in Klaten, Indonesia.This data set is split into training data 60% and testing data 40%. Meanwhile, statistical analysis considers the significant difference between the proposed models. Experimental results show that NNBP provides smaller RMSE than an SVM model.The proposed model contributes not only to update the original instrument, but also applicable and beneficial for the industry, particularly in deciding effective inventory replenishment decision on production quantity.