Building quality complaint text classification method and system

The invention discloses a method and a system for classifying building quality complaint text, the method comprises data collection and analysis: collecting existing building quality complaint text and manually matching a complaint category to each complaint; pre-processing the complaint text: useing the word segmentation software to deal with the word segmentation of the complaint text and deletethe stop word; model construction: constructing a text classification model based on a convolution neural network; model optimization: carrying out training on and verification of the model using themarked text data, automatically adjusting the model parameters by comparing the accuracy of the verification results of different batches and finally obtaining the optimal classification model. As the invented technical scheme of the building quality complaint text classification method and system adopts the machine learning based on the convolution neural network, the complaint text is expressedas a space vector containing semantic information, complaint text can be efficiently and accurately classified, avoiding the understanding deviation caused by text language expression under manual classification mode, which is beneficial to the follow-up processing of the building quality complaint text.