Classification of Crude Palm Oil Quality Using K- Nearest Neighbor (K-NN)

The value of agricultural exports in Indonesia is mostly contributed by crude palm oil (CPO). The productivity of the best quality crude palm oil is very important to consider to increase Indonesia's competitiveness in the global market. However, determining the quality of CPO is susceptible to subjectivity from one or more parties, so it requires an objective approach with a machine learning algorithm. This study applies K-Nearest Neighbor (K-NN) to classify CPO quality using data from 2019 to 2023, totaling 6,790 data. The parameters for determining CPO quality are based on the percentage of FF A, moisture, dirt, and DOBI contained in CPO. The study results showed that the K-NN model with a scheme of 90 training data and ten testing data using SMOTE, with a K value = 5, and the Manhattan distance calculation is the most optimal model with an average accuracy of 99.54%. Furthermore, the confusion matrix results showed an accuracy value of 99.68%, precision of 98.92%, recall of 100%, and F1-score of 99.45%. From these values, the K-NN method can classify CPO quality quite accurately.