Abstract
Currently, researchers in various fields are using fuzzy discretization for decision-making. Discretization construction for numerical data involving a combination of fuzzy membership functions is significant because it can affect the performance of the model used. Classification of AISI 1045 surface roughness with satisfactory performance is needed to improve efficiency and extend the service life of AISI 1045 products. This study utilizes three fuzzy membership functions: beta, sigmoid, and linear in constructing fuzzy discretization on machining factors and tangential roughness levels to classify the axial roughness level of AISI 1045. Classification is performed using an ensemble of single naïve Bayes methods integrated with fuzzy discretization. These single methods are distinguished based on the combination of fuzzy membership functions used in the discretization. The results of the study show that the integration of fuzzy discretization through a combination of fuzzy membership functions, namely beta, sigmoid, linear, and fellow beta functions in the MNB method provides different performance, even the performance of MNB with fuzzy discretization using a combination of beta and sigmoid is almost the same or not statistically significantly different from the performance of the ensemble method. However, the ensemble method built provides the best performance for classifying the surface roughness level of AISI 1045, with Accuracy, Precision, Recall, F1-score, AUC, Balanced Accuracy, and G-Mean of 85.42%, 55.33%, 73.14%, 62.63%, 71.71%, 81.71%, and 81.04%, respectively.