The Text Feature Selection Research Based On Auto-encoder Neural Network

The significant target,been based on auto-encoded neural network text feature extraction,is to improve the efficiency of current text mining.Due to the rapid development of the network,a large amount of information,especially the major text information spread in the network,and make the traditional text mining technology defects exposed.In order to deal with a large amount of text data,the auto-encoded neural network which has the advantage of reducing the large amount of text features,does not damage to its accuracy and increase its speed.It can meet the high efficiency of modern network.In this paper,auto-encoded the neural network is one of methods of deep learning,which is a new direction in the research of machine learning.The idea of deep learning comes from the study of artificial neural networks,which is a structure with a multi-layer perceptron.Deep learning uses a combination of low level features to synthesize more abstract high-level features to express,and its purpose is to discover the information of the distributed characteristics of data.The auto-encoded algorithm is a kind of unsupervised learning algorithm based on deep learning.It uses the multi-layer perceptron structure and the back propagation algorithm of neural network in deep learning.The two significant features of auto-encoded the neural network is that the number of output nodes of the neural network is equal to the input nodes and the number of nodes in the hidden layer is less than the input and output nodes.The dimension of the text feature dimension is reduced,and the computation quantity of space complexity is also reduced.With this method,the high efficieny of auto-encoded the neural network is realized.In the paper,the 50 groups of similar information regard as an example(each group is not more than 30 Chinese characters)which are transformed into the dot matrix code,and 35 groups of examples choosing from 50 groups are a trainning sample for encoding and are also used as the input information of the neural network under the condition of combining with the visual technology.The study sample is trained by using neural network model of MATLAB tool to carry on the experiment.Through the model deal with the input 35 groups of samples,the method of feature transformation is that sending the original data samples which is from the original feature space map,into a new feature space,and the text feature library is established at the same time.The remaining 15 groups use the method for text feature extraction,and the feature compare with the text feature which is in the database.Experiments show that the auto-code neural network can get good results that the method is good at classifying the text,and can reduce the computational complexity in time and space.