Complex-Valued Neural Network And Fuzzy Inference System For Image Diagnosis Of Rice Leaf Diseases

Rice serves as a crucial food crop and holds significant importance in Indonesia's agricultural sector. so the health of rice leaves determines the productivity of the crop. Serious problems such as crop failure often occur due to leaf disease attacks caused by pests or unfavorable climatic factors. Controlling these diseases requires proper knowledge so as not to cause negative impacts on the ecosystem due to misdiagnosis. This research develops a Complex-Valued Neural Network (CVNN) and Fuzzy Inference System (FIS) based method to identify the type of disease and determine its severity. CVNN was used to classify leaf images based on detected visual traits, while FIS analyzed the relationship between these traits and disease severity using fuzzy rules constructed from expert data or input. The results show that CVNN provides superior performance compared to CNN, CVNN model with an accuracy of 92%, where all classes produce high and balanced. While the CNN model also provides satisfactory results with an accuracy of 89%, although there is still an imbalance in some classes. The results of the FIS model on the image The severity of the image of rice leaf disease is the most high category in the leaf blast class is the highest of all classes. The combination of CVNN and FIS model proves that this hybrid approach is effective to support diagnosis, so it can help farmers in making early and precise decisions.