XGBoost-Based Prediction Model for Mine Signal Classification

With the continuous development of coal industry mining, the risk of underground shockwave disaster is increasing, and in order to reduce the loss of personnel and property caused by shockwave, the accurate classification and prediction of the detected signals has become an important research topic. In this paper, to address this problem, based on the idea of time series window, we construct the feature engineering for monitoring acoustic emission (AE) and electromagnetic radiation (EMR) signals respectively according to relevant data, and use model algorithms such as XGBoost and Particle Swarm Optimization (PSO) to train and optimize the parameters. For the selection of signal features, we rank the importance of features based on the SHAP method, and finally derive the features of the signals under different categories, and apply these features to classify and identify and predict the signals from related data. This research provides powerful tools and methods to cope with the classification and identification of signals in different situations.