An ontology-based agriculture decision-support system with an evidence-based explanation model

Effective management of plant diseases and pests requires knowledge that covers multiple domains. At the same time, retrieving the relevant information in a timely manner is always challenging, due to the unstructured nature of agricultural data. Over the years, efforts have been made to develop an ontology-based Decision-Support System (DSS) to facilitate the diagnosis and control of plant diseases. Some major issues with these systems are that: (1) they do not adopt the full extent of the ontological constructs to represent domain entities, which, in turn, reduces reasoning capabilities and prevents systems from being more intelligent, (2) they do not adequately provide the desired level of knowledge to support complex decisions, which requires many factors to be considered, (3) they do not adequately explain or provide evidence to demonstrate the validity of the system's outputs. To address these limitations, we present a novel system termed Agriculture Ontology Based Decision Support System (AgrODSS), which aims to assist in plant disease and pest identification and control. AgrODSS architecture consists of two semantic-based models. First, we developed Plant Diseases and Pests Ontology (PDP-O) to capture, model, and represent diseases and pest knowledge in a machine-understandable format. Second, we designed and developed an Evidence-Based Explanation Model (EBEM) that points to related evidence from the literature to demonstrate the validity of the system outputs. We demonstrate the effectiveness of AgrODSS by executing various queries via AgrODSS SPARQL Endpoint and obtaining valuable information to support decision-making. Finally, we evaluated AgrODSS practically with domain experts (including entomologists and pathologists) and it produced similar answers to those given by the experts, with an overall accuracy of 80.66%. These results demonstrate AgrODSS's ability to assist agricultural stakeholders in making proper disease or pest diagnoses and choosing the appropriate control methods.