JoWUA

Abstract

Based on the Indonesian National Standard (SNI) 01-3392-1994, clove quality in Indonesia is categorized into three distinct classes (Quality 1, 2, and 3). However, quality determination presents significant challenges due to the similar morphological characteristics across classes and limited available data, necessitating manual feature extraction. This manual process often yields irrelevant features, highlighting the need for robust feature selection methods. This study demonstrates that ensemble feature selection significantly enhances the performance of conventional machine learning models (K-NN and Naïve Bayes) in SNI-compliant clove quality classification. The proposed approach employs five distinct feature selection methods to assign importance scores to features, with the final selected features representing those consistently identified across all methods. Experimental results indicate notable improvements, including a 2.66% accuracy increase in K-NN classification and reduced computational time across all tested models. These findings substantiate the effectiveness of ensemble feature selection for optimizing conventional machine learning applications in agricultural quality assessment.