A uniform manifold approximation and projection (UMAP) and long short-term memory (LSTM) deep learning model have been proposed to forecast a sealed-off area's fire status in underground coal mines. It protects miners' life by providing early warning to the miners regarding the impending mine hazards. The proposed forecasting model graphically displays fire status in the form of Ellicott's extension graph. An experiment has been conducted to measure the proposed forecasting model's efficiency and two existing machine learning models, namely support vector regression (SVR) and auto-regressive integrated moving average (ARIMA) models. It has been found that gas concentration prediction of the proposed UMAP-LSTM model has the lowest root mean square error of 0.288, 0.006, 0.0995, 0.902, 0.238, 0.452, and 0.006 for O₂, CO, CH₄, CO₂, H₂, N₂, and C₂H₄ gases respectively than the existing SVR and ARIMA models, which indicates higher efficiency of the proposed prediction model.

提出了统一的流形逼近和投影（UMAP）和长短期记忆（LSTM）深度学习模型，以预测地下煤矿的禁区火灾状况。它通过向矿工提供有关即将发生的地雷危害的预警来保护矿工的生命。拟议的预测模型以Ellicott扩展图的形式以图形方式显示火灾状况。已经进行了一项实验，以测量建议的预测模型的效率以及两个现有的机器学习模型，即支持向量回归（SVR）和自回归综合移动平均（ARIMA）模型。已经发现，所提出的UMAP-LSTM模型的气体浓度预测的最低均方根误差为0.288、0.006、0.0995、0.902、0.238、0.452和0。，CO，CH ₄，CO ₂，H ₂，N ₂和C ₂ H ₄气体分别比现有的SVR和ARIMA模型高，这表明所提出的预测模型具有更高的效率。