Prediction of Manufacturing Processes Errors: Gradient Boosted Trees Versus Deep Neural Networks

Abstract

In this paper we investigate the use of machine learning techniques for optimizing manufacturing processes operation. More precisely we propose, compare and contrast two approaches for predicting errors in manufacturing processes. The first approach is based on machine learning algorithms while the second one uses deep learning techniques. Both approaches are validated using a dataset from literature, the SECOM dataset, which is representative for manufacturing processes. For the machine learning approach features are selected using the Multivariate Adaptive Regression Splines (MARS) algorithm and data is classified using the Gradient Boosted Trees (GBT) algorithm, while for the deep learning approach features are selected using a Support Vector Machine (SVM) algorithm and data is predicted using a Neural Network (NN). The evaluation results show that the best results are obtained using the deep learning approach.