Classification based on K-Nearest Neighbor and Logistic Regression method of coffee using Electronic Nose

Abstract

Coffee has its own scent of identity which can be felt directly with the ability of the human sense of smell. With a specific coffee aroma that can be used to identify the type of coffee. In this study we propose that E-Nose (Electronic Nose) can be used to identify coffee based on the aroma of coffee converted into value data used for the classification process. The initial step is the data validation process using the calculation of the average value, standard deviation, Minmax. After conducting the dataset validation process, the next step is to implement the Logistic Regression (LR) and K-Nearest Neighbor (KNN) classification methods. The accuracy value is derived from the Confusion Matrix evaluation method, TP, TN, FP and FN values. This study focuses on finding the best classification accuracy value with the criteria having the highest accuracy value. This system can be used to classify types of coffee with a mixture of coffee and milk. This study will compare the results of classification using the two classification methods. Based on the results of the accuracy of the two methods presented the best results using the KNN method with a statistical calculation is 97.7%.