Comparison of Tree Method, Support Vector Machine, Naïve Bayes, and Logistic Regression on Coffee Bean Image

Abstract

Coffee is one of the many favorite drinks of Indonesians. In Indonesia there are 2 types of coffee, namely Arabica & Robusta. The classification of coffee beans is usually done in a traditional way & depends on the human senses. However, the human senses are often inconsistent, because it depends on the mental or physical condition in question at that time, and only qualitative measures can be determined. In this study, to classify coffee beans is done by digital image processing. The parameters used are texture analysis using the Gray Level Coocurrence Matrix (GLCM) method with 4 features, namely Energy, Correlation, Homogeneity & Contrast. For feature extraction using a classification algorithm, namely Naïve Bayes, Tree, Support Vector Machine (SVM) and Logistic Regression. The evaluation of the coffee bean classification model uses the following parameters: AUC, F1, CA, precision & recall. The dataset used is 29 images of Arabica coffee beans and 29 images of Robusta beans. To test the accuracy of the model using Cross Validation. The results obtained will be evaluated using the confusion Matrix. Based on the results of testing and evaluation of the model, it is obtained that the SVM method is the best with the value of AUC = 1, CA = 0.983, F1 = 0.983, Precision = 0.983 and Recall = 0.983.