Selection of Superior Rice Seeds by Applying Fuzzy Method and K-Means Clustering

Abstract

This study aims to develop a superior rice seed selection method using a Fuzzy and K-Means Clustering approach, with a case study in Kutai Kartanegara Regency, East Kalimantan Province, one of Indonesia's major rice-producing regions. The Fuzzy method is used to handle uncertainties in assessing seed characteristics, allowing each seed attribute (such as plant height, amylose content, grain weight, and yield) to have a membership value within specific categories. This fuzzification process provides flexibility in evaluating seed quality in stages, which is then converted through defuzzification to obtain a final score determining seed quality. K-Means Clustering plays a role in grouping seeds based on characteristics that have been assigned membership values. This algorithm divides seed data into several clusters, such as low, medium, and high quality, by calculating the distance between seed characteristics and each cluster's centroid. This iterative process yields seed groups with similar characteristics, simplifying recommendations for superior varieties. The evaluation was conducted using clustering accuracy metrics and silhouette score validation to ensure cluster cohesion and separation. The study results demonstrate that this method effectively identifies high-quality rice seeds with high accuracy. Recommended varieties include standard rice seeds like Mengkongga and Ciherang, as well as superior varieties like Inpari 32, Inpari 48, Padjajaran Agritan, Inpari IR Nutri Zinc, and Pamera, which are well-suited to Kutai Kartanegara’s specific conditions. Implementing this method is expected to assist farmers in selecting high-quality seeds, thereby supporting increased crop productivity in the study area.