Classification technique of chinese agricultural text information based on SVM

In order to provide personalized services for agricultural information recommendation, it was needed to organize and classify information efficiently. According to the characteristics of agricultural texts, a Chinese agricultural text classification model was proposed based on linear support vector machine (SVM). Firstly, an agriculture-domain-based dictionary was built. Secondly, a feature vector was extracted and the weight for each feature in a vector was selected. Lastly, a text classification model was established. The model was tested on 1071 documents which were belonged to four classes: planting, forestry, animal husbandry and fisheries. The results showed that the accuracy was 96.5% and the recall rate was 96.4%. Both of their performances were higher than those of the ones using other classification methods, such as the Bayesian, decision tree, KNN, SMO algorithm and neural network. The model was applied to the platform for agricultural internet of things (IOT) industry integrated information service. The performance showed that the method can automatically classify Chinese agricultural text information and the response time met the system requirements. © 2015, Chinese Society for Agricultural Machinery. All right reserved.