IMPROVING PERFORMANCE OF COPRA TYPE CLASSIFICATION USING FEATURE EXTRACTION WITH K-NEAREST NEIGHBOUR ALGORITHM

Abstract

Through observations and interviews with coconut farmers in the Indragiri Hilir, Bitung, and Palembang districts, it has been found that the assessment of copra quality continues to encounter various obstacles, such as demanding substantial labour, time, and expenses. This research focuses on categorising copra types by extracting features and utilising the K-nearest neighbour algorithm. The research involves a dataset from copra warehouses in Indragiri Hilir Regency, Riau Province, consisting of 613 digital images classified into three categories: edible, regular, and reject. When categorising copra types, feature extraction methods like colour, shape, and texture features are taken into account. According to the study findings, the model accuracy when all features are taken into account is 84%.