Coal Mine Risk Classification Prediction Model Based on BERT

This paper proposes a multi-label classification model to predict various risks that may occur in coal mines. A Lite BERT(ALBERT) is used as the upstream model to extract text features, and TextCNN is used to realize the downstream classification task. The model significantly reduces the training cost without affecting the classification effect, and the final prediction accuracy reaches 98 \(\%\) , which can meet the requirements of system design. As a lightweight coal mine risk classification and prediction model, it has a small number of parameters and high prediction accuracy, which has certain practical value and can be used to help managers to make risk assessment and decision-making.