Classification of South Sulawesi Rice Quality Based on Website Using Machine Learning

Abstract

Rice is a staple food commodity in Asian countries like Indonesia, but there are still challenges in quality assurance and standardization. The objectives of this research are to design and to develop a website-based system to classify the quality of rice using machine learning methods, to analyze the performance of the website-based South Sulawesi rice quality classification system, and to determine the accuracy of the classification system. This research uses the Convolutional Neural Network (CNN) method to classify rice quality based on some characteristics such as texture, shape, color, and size. The research succeeded in designing and developing a website-based system called "BERAS PINTAR". This system integrates a CNN model to classify 10 types of rice and three rice quality categories: premium, medium, and regular rice, based on input images of rice. Classification performance analysis shows very effective results. The model achieved an accuracy of 99.05% and was able to process classification quickly, proving its performance is satisfactory and functional. The resulting system proved accurate and efficient in identifying rice quality. The accuracy test reached a maximum of 100% and confidence levels ranging from 60% to 100%. This level of accuracy is consistent for rice varieties with distinctive visual characteristics.