D-P-Transformer: A Distilling and Probsparse Self-Attention Rockburst Prediction Method

Rockburst may cause damage to engineering equipment, disrupt construction progress, and endanger human life. To this day, the occurrence of rockburst remains complex difficult predict. This study proposes D-P-Transformer algorithm address issue by improving embedding structure Transformer for specific applications data. reduce computational requirement, sparse self-attention is adopted replace self-attention. A distilling operation multiple layer replicas are simultaneously used enhance robustness speed up algorithm’s process. Taking all relevant factors into consideration, experiments conducted on seven large-scale datasets with different training ratios verify reliability proposed prediction algorithm. As compared original algorithm, shows average reductions 24.45%, 46.56%, 17.32%, 48.11% in mean absolute error (MAE), square (MSE), root (RMSE), percentage (MAPE), respectively. The results indicate that novel superior could be coal mine analysis.