Klasifikasi Menentukan Kualitas Sarang Burung Walet Menggunakan Metode Extreme Gradient Boosting (XGBOOST) Berdasarkan Citra

Abstract

Edible bird's nest is a high-value export commodity whose market price heavily depends on physical qualities such as color, cleanliness, and cup shape. Conventional quality determination in the field still relies on manual visual observations, which are highly susceptible to subjectivity, human error, and prolonged operational time. This study proposes an automated classification system framework using digital image processing and the Extreme Gradient Boosting (XGBoost) algorithm to evaluate the quality standards of edible bird's nests objectively and non-destructively. The computational feature extraction focused on four main parameters: color (mean grayscale), shape (circularity), structural integrity (solidity), and cleanliness (purity). Utilizing a dataset of 1,005 images categorized into Premium, Standard, and Low classes, the quantitative performance of the XGBoost model was evaluated. The testing results demonstrated a flawless training accuracy of 100.00% and an overall test accuracy of 80.60%. The algorithm proved highly sensitive in identifying the Standard class with a recall value of 0.86, and robust in isolating the Low class with a precision of 0.82. Comprehensively, this hybrid approach integrating image processing and XGBoost successfully minimizes human subjectivity and provides a precise quality standardization solution for the edible bird's nest industry.