The Research And Application Of Short Text Classification Based On Machine Learning In Nuclear Power Quality Management

In today’s information age,enterprises generally pursue the fine and efficient transaction processing.The management of the manufacturing enterprises,R&D enterprises,even nuclear engineering company are hoping to carry out the classification of management event text.This project is to use the machine-learning algorithms to solve the classification problem in the specified process,which is from the manufacturing stage to the service stage,according to the management requirement of nuclear power experience feedback(EF).The classification content is from equipment quality managent corpus which is collected from the process.First,the corpus of quality management of NPP equipment is prepared as the train set with artificially marked.Then,Labeled-LDA,SVM and Convolutional Neural Network are used to train the corpus.The test set of the project is applied to compare the effects of the classification.Finally,the paper analyses the classification results.By using machine-based learning techniques,nuclear power engineering companies can efficiently classify quality events and liberate managers from complex manual classification tasks,given the lack of quality management personnel and the large number of classified events.At the same time,the classification of nuclear energy events is also the basic task required by the International Atomic Energy Agency(IAEA)for the management of nuclear power experience feedback.By transforming unstructured event texts into statistically available structured data,a solid data base is provided for experience feedback and knowledge management of nuclear engineering.