Prediction and Early Warning Method for Flea Beetle Based on Semi-supervised Learning Algorithm

The prediction and early warning for vegetable crop diseases and insect pests commonly based on expertspsila knowledge of plant protection while math modeling methods are used to analyze the associated data quantitatively scarcely. This paper takes advantage of labeled historical data to extract association rules as the supervised information, and combining with the use of the unsupervised learning method iterative self- organizing data analysis techniques algorithm (ISODATA) to establish the model semi-supervised learning algorithm for predicting and early warning vegetable pest flea beetle. Semi-supervised learning algorithm not only obtains a high accuracy of supervised learning method, but also takes advantage of the flexibility of unsupervised learning method, which is significative for both research and practice. The experimental results of Guangdong vegetable pest flea beetle show the prediction and early warning accuracy of semi-supervised learning method provides a higher accuracy rate than that of k-mean clustering, support vector machine and RBF neural network in the same condition.