Research On Text Classification Based On Hybrid Model Of Deep Learning

With the rapid development of the Internet, text classification has become a key technology under the age of big data. Text information contains a lot of valuable information. How to effectively manage these text information and effectively obtain valuable information has become a challenge in the field of information science. Text classification is the key point of text information processing. It plays an important role in text information processing.At present, deep learning has been widely used in handwriting recognition,image recognition, voice recognition and other fields, but the application of text classification is still relatively small. This paper makes full use of the good ability of deep learning which can learn characteristics. We propose a hybrid model based on deep learning and design a text classifier based on the hybrid model. The hybrid model uses a sparse auto encoder and a deep belief network. Those model are two kinds of common deep learning model. The hybrid model has three components:a two layer using sparse auto encoder to construct, a three layer using deep belief networks to construct and a classification layer using softmax regression.In order to test the classification performance of the classifier based on the hybrid model of deep learning, the related experiments were conducted on the English data set 20 Newsgroup and Chinese data set Chinese corpus of Fudan University. In the experiment of English text classification, the classifier based on hybrid model of deep learning obtained perfect classification accuracy. To further verify the superiority of its performance, we performed a experiment with Naive Bayes classifier, KNN classifier and support vector machine classifier. The classifier based on hybrid model of deep learning is slightly better than Naive Bayes classifier, KNN classifier and support vector machine classifier. In the experiment of Chinese text classification, Chinese corpus of Fudan University experimentally obtained good classification results, and we discuss the influence of different parameters on the classification accuracy.