Prediction of mining induced subsidence by sparrow search algorithm with extreme gradient boosting and TOPSIS method

Land subsidence caused by coal mining has caused damage to villages, buildings and farmland, threatening the human settlements and ecological environment of mining areas. In this work, a novel intelligent approach for predicting mining land subsidence was developed. The novel sparrow search algorithm combined with efficient extreme gradient boosting and technique for order preference by similarity to an ideal solution were applied to achieve this goal. 140 sets of mining land subsidence data composed of 6 main parameters were selected as input independent variables with the maximum subsidence value as the output-dependent variable. The sparrow search algorithm was used to optimize the hyperparameters of extreme gradient boosting, and the technique for order preference similarity to an ideal solution method was used to select the best prediction model. Finally, five machine learning models such as extreme gradient boosting, random forest, adaptive boosting, bootstrap aggregating, and gradient boosting were compared to the proposed model. 140 data samples are retrieved from the literature and the determination coefficient, root mean square error, mean absolute error, and the variance accounted for were selected to evaluate the model performance. In this work, the Shapley additive explanations method was used to interpret the importance of features and their contribution to the maximum subsidence prediction. Compared with other machine learning models, the model proposed in the research is more reliable. The research results show that compared with other machine learning models, the proposed is the most reliable one. This study can aid miners in land subsidence prediction.