Cucumber diseases diagnosis based on multi-class SVM and electronic medical record

Cucumber is one of the most popular vegetable varieties, but leaf disease of cucumber is the key factor restricting the increase of yield. Common cucumber diseases include downy mildew, powdery mildew and gray mold. The plant electronic medical records (PEMRs) formed by “plant clinic” are the diagnosis record of the real disease occurrence issued by the plant doctor, which provides a new idea for the cucumber disease diagnosis. The efficient mining of prescription big data to facilitate precise diagnoses of crop pests and diseases represents an emerging challenge. The data mining technology represented by machine learning has attracted wide attention. Therefore, 15 diagnosis models are proposed to deal with this problem. Since SVM has many advantages including implementing the structural risk minimization principle and can effectively deal with the small sample data, five algorithms based on SVM have achieved better diagnosis performance in comparison with others. Moreover, the highest prediction accuracy is beyond 80%. In addition, the prediction performance has been improved after employing the undersampling technology for the imbalanced data. This means they are suitable for this cucumber diseases diagnosis.