Abstract

Cassava (Manihot esculenta Crantz) is a vital staple food and industrial raw material, particularly in Indonesia, where it holds significant economic importance. Despite its adaptability, cassava cultivation faces serious threats from plant diseases, leading to reduced productivity and quality. Traditional agricultural methods, often lacking modern technological interventions, exacerbate these challenges. To address this, a web-based expert system was developed to assist farmers in diagnosing and managing cassava diseases. The system integrates the Certainty Factor (CF) and Dempster-Shafer (DS) methods to handle uncertainty in disease diagnosis. Using the Adira 1 cassava variety, data was collected from a 2-hectare area in Bogor, focusing on five types of cassava diseases: brown leaf spot, blight leaf spot, bacterial blight, anthracnose, and root rot.

The CF method quantifies the certainty of specific facts, while the DS method combines evidence from various sources to determine the belief level of potential diagnoses. A comparative analysis of these models was conducted using 20 case studies. The results showed that the CF model achieved a 95% accuracy rate, slightly outperforming the DS model's 90% accuracy. Additionally, the system's usability was evaluated using the USE Questionnaire, which rated it as "Highly Suitable" with an overall score of 84.5%.

This study demonstrates that while both CF and DS models are effective in diagnosing cassava diseases, the CF model offers slightly higher accuracy. The developed expert system is not only accurate but also user-friendly and accessible, making it a valuable tool for improving cassava disease management practices.

Keywords

Cassava, Disease Diagnosis, Certainty Factor, Dempster-Shafer, Expert System, Agriculture Technology.