Incorporating multi-features and XGBoost algorithm for gas concentration prediction

Gas concentration accidents occur frequently under the coal mining scenario. Therefore, predicting gas concentration is of great significance for preventing accidents. Although previous studies have conducted extensive work on gas concentration prediction, the gas concentration is affected by a variety of complex factors under the coal mining scenario, showing instability and nonlinearity, and bringing limitations for accurate prediction. Recently, researchers have pay attention to deep learning methods due to their good performance. However, existing methods require more training time as the amount of data increases. Moreover, most of them only considered historical gas concentrations as the single input. To solve the above problems, the paper proposes a multi-features and XGBoost algorithm gas concentration prediction model. This model can simultaneously incorporate multi-features, including historical gas concentration values, temperature, wind speed, and other characteristics as the input, while the inner gradient boosting algorithm and decision tree can improve the training speed of the model. Extensive experiments proving that our method can achieve better performance and faster speeds than existing advanced neural network models.