Prediction of Coffee Bean Quality Using Segmentation Methods And K-Nearest Neighbor

Abstract

The condition of people's coffee farming management is relatively poor when compared to large stateowned plantations. The main problem in smallholder plantations is the quality of the results that do not meet standardization. This study designs a system that is able to identify the quality of coffee beans using Segmentation, K-Nearest Neighbor and Gray Level Co-occurrence Matrix methods. Based on the test results using texture feature extraction, the highest accuracy was obtained at K-5 of 85%. It is possible that if the K value used is too small, there will be a lot of noise which reduces the level of accuracy in data classification, but if the K value is too large it can cause errors in the range of values taken, which will indirectly affect the level of accuracy. The results of the study were the identification of coffee beans with good quality or poor quality. It is hoped that this research can contribute to improving the quality of people's coffee so that it can increase the production of people's coffee that is able to compete in the market.

Keywords—Gray Level Co-occurrence Matrix, K-Nearest Neighbor, Segmentation