METODE KLASIFIKASI MUTU GREENBEAN KOPI ARABIKA LANANG DAN BIASA MENGGUNAKAN K-NEAREST NEIGHBOR BERDASARKAN BENTUK

Abstract

During this time, the Greenbean coffee sorting process is still done manually which still has many shortcomings. Manually, this result is classified in inappropriate and inconsistent classification results due to human negligence. Grading in the processing and marketing sectors is important. Inappropriate grading opposes farmers simply because Lanang and ordinary Arabica coffee are the same. Hence, we need a consistent classification system. This research uses image processing to recognize Greenbean Arabica coffee. K-NN (K-Nearest Neighbor) method is used for a quality classification. This research will classify Arabica Greenbean coffee into 4 quality classes, namely intact Lanang Arabica, broken Lanang Arabica, intact ordinary Arabica, and ordinary broken Arabica. The search of trial process shows that K-NN classification feature is able to recognize Arabica coffee Greenbean into 4 classes with an accuracy value of 63.5%, very good at recognizing 90% of regular Arabica intact and 97% of whole Arabica intact. However, it is still weak in recognizing broken coffee Greenbean based on its type. The area feature is better in recognizing Arabica coffee Greenbean based on 4 classes with an accuracy of 69.8%. This research obtains 120 datasets from 80 tested data trains and 40 tested random data.