Pengembangan Sistem Pakar Hama Penyakit Tembakau Berbasis Web dengan Mengintegrasikan Graf Pengetahuan dan Bayesian Network

Abstract

Indonesia is a nation where agriculture plays a vital role, contributing significantly to both gross domestic product and employment opportunities. Tobacco stands as one of the main plantation commodities, with national production reaching 353,39 thousand tons per year in 2024, of which East Java accounts for 213,57 thousand tons. Despite the impressive output, pest and disease attacks remain a major challenge, increasing the potential failure by approximately 10–15%. Early identification of pests and diseases through observation of physical symptoms is crucial to prevent production losses and can be used as evidence in an expert system. This research develops a web-based expert system for diagnosing tobacco pests and diseases by integrating knowledge graph and Bayesian Network. Knowledge graph is used to systematically represent knowledge about pests, diseases, and their symptoms, while Bayesian Network enables probabilistic inference. The integration of these two methods aims to produce an accurate diagnostic system with clear chains of evidence, measurable confidence levels, and a web interface that is easily accessible to farmers directly in agricultural fields. This system is expected to reduce crop failure rates and economic losses due to tobacco pest and disease attacks while supporting better decision-making in plant pest and disease management.