Evaluation and prediction of coal seam mining mode: Coefficient of Variation-TOPSIS and CNN-NGO methods

Selecting the appropriate method for coal seam mining is the foremost task for the safe and efficient production of coal mines. This study explores and validates an integrated evaluation system that enhances the accuracy of predicting coal seam mining mode by comparing traditional evaluation methods with machine-learning techniques. The weights of the evaluation indicators for coal seam mining were allocated using the coefficient of variation method, followed by a comprehensive evaluation using the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). In addition, a Convolutional Neural Network (CNN) regression model was constructed and optimized with the Northern Goshawk Optimization (NGO) algorithm, resulting in a more precise CNN-NGO prediction model. A detailed comparative analysis of these two models was conducted. The results indicate that the level of equipment is a key factor affecting the method of coal seam mining, holding the highest weight among all evaluation indicators. The CNN-NGO model demonstrates excellent performance in predicting coal seam mining mode, with its predictions highly consistent with actual mining practices. Specifically, in the practical application case of the WJZ 15206 working face, the model successfully predicted the high mining height as the most likely mining method, with a prediction index of 2.4265, close to the normalized output value of 2 for large mining height. This study not only provides scientific methods and tools for the evaluation and prediction of coal seam mining mode but also contributes to the intelligent development of the coal mining industry.