Research On Agricultural Information Classification Method Based On Deep Learning

With the country’s strong support for the three rural issues and the rapid development of Internet technology,agricultural information continues to expand the expansion of agricultural information development is rapid,online agricultural information documents have been quantified.How to achieve rapid search of agricultural information from the quantitative agricultural information,accurate positioning had become increasingly difficult.Under this background,it is very important to choose the optimized method of agricultural information text classification to assist in the quick retrieval of agricultural information text.In this paper,the classification method based on decision tree,Bayesian and deep learning is studied,and the network structure and network training process of CNN model in deep learning are discussed.The automatic classification of agricultural information text is realized and text classification is improved The accuracy and efficiency,thus increasing the value of the use of information.The main work is as follows:(1)Data acquisition and pre-processing part.The paper uses the crawler tool to obtain the document from the China Agricultural Information Network as the agricultural information data set,and then uses the Jieba word and Pynlpir two word segmentation methods to segment the data set,and use the stop word list to remove the word in the word file,the number And then use the commonly used feature selection evaluation function to select the feature,on the basis of this proved that the use of convolution neural network model automatically extract the feasibility of agricultural information characteristics.(2)Vectorization of two methods of agricultural information text.One is the Chinese word segmentation,to stop the word after the extraction of the text feature words directly expressed as a text vector method;one is the Chinese word segmentation,to stop the word directly after the word vector method;the use of word vector method to avoid the traditional vector representation dimension too high the problem,the use of the deep learning methods can automatically extract the agricultural information text feature words.(3)Based on the vector file generated by preprocessing,the classification of agricultural information text is realized by using the convolution neural network model of decision tree,Bayesian and deep learning,and the results of the operation are analyzed theoretically.And then the use of clustering method to verify the distribution of the data set category and use the pie chart to visualize the display,to verify the classification of the two categories and the results of the difference is because the data set category Include the number of documents caused by the imbalance.The feasibility of the application of convolution neural network in agricultural information text classification is verified by experiments.Compared with other existing classifiers,the advantages of convolution neural network model in agricultural information classification are analyzed.(4)The optimization model of the convolution neural network model for agricultural information classification is put forward and the experiment is carried out.The experimental results are compared and analyzed.The experimental results show that the nodes in the network structure classified by agricultural information are Sigmoid Function when the network classification performance decreased significantly,and each node using Relu Function when the network classification performance significantly improved.In the experiment of adjusting the number of convolution cores,the number of convolution Core in the network model is increased to twice that of the original network,and the network finally achieves the accuracy rate of 99.40%.