Automatic Classification of User Requirements Information Based on Convolutional Neural Network

User requirements information is semi-structured or unstructured data with sparse features and short lengths. Since manual labeling methods nowadays are time-consuming, which cannot meet increasing requirements for classifying massive user requirements information, an automatic classification method of user requirements information based on Convolutional Neural Networks (CNN) is proposed. Firstly, NLPIR word segmentation tool is used to preprocess word segmentation and remove stop words in user requirements information. Secondly the word vector is trained by word vector model (word2vec). Thirdly convolutional neural network is used to extract abstract features of text information. And finally these features are used as input,and the softmax classifier is used to achieve automatic classification of user requirements information. Compared with the traditional classification method based on probability and statistics, the classification accuracy of the automatic classification model of user requirements information based on CNN has been improved by 13.54%. The experimental results show that the CNN classification algorithm is more suitable for the automatic classification of the user requirements information.