Predicting Safety Logic Device Solutions via Decision Trees and Rules Algorithms

Considering the extensive data sets and statistical techniques, a digital factory (plant) embodies a branch of machine learning that has an impact on machine safety. We propose a study based on an application of decision trees and rules algorithms (JRIP, J48, Random Forest, Random Tree, and PART). Our experimental data were collected from various industrial machine safety solutions. Diverse validation techniques were employed to derive the classification performance of each method; the approach is expected to simplify the user choice of a suitable safety logic device type. In this study, the overall classification methods proportion of individual safety logic device solutions were correctly assigned by using the training-evaluated test mode, and the prediction accuracy reached 100%; further, when assessing the 5-fold cross-validation test mode, we obtained the success rate of 82% (JRIP and PART). PART as the best method was correctly assigned for the 10-fold cross-validation test mode (85%). New developments within the broad province of machine learning, including the concepts characterized in our study, may facilitate effective assessment of machine safety systems.