煤矿安全隐患信息自动分类方法

Automatic classification method of coal mine safety hidden danger information

人工分类方式难以满足海量煤矿安全隐患信息的分类要求,而基于概率统计的文本自动分类方法分类准确率较低。针对上述问题,提出了一种基于Word2vec和卷积神经网络的煤矿安全隐患信息自动分类方法。首先对隐患信息进行分词、去停用词等预处理,然后应用Word2vec来表征词之间的语义相似性关系,最后利用卷积神经网络提取隐患信息的局部上下文高层特征,并使用Softmax分类器实现隐患信息的自动分类。实验结果表明,该方法实现了端到端的自动分类,可有效提升分类的准确性和全面性。

Manual classification method is difficult to meet classification requirements of massive coal mine safety hidden danger information,and automatic text classification method based on probability statistics has low classification accuracy rate.In view of the above problems,an automatic classification method of coal mine safety hidden danger information was proposed which was based on Word2vec and convolutional neural network.Firstly,hidden danger information is pre-processed through word segmentation and stop word deletion.Then semantic similarity between words is represented by employing Word2vec.Finally,local context high-level features of hidden danger information are extracted by use of convolutional neural network,and Softmax classifier is used to realize automatic classification of hidden danger information.The experimental results show that the method realizes end-to-end automatic classification and can effectively improve accuracy and comprehensiveness of classification.