Abstract: To address the low precision and extensive control in coal mine subsidence management, this paper proposes an intelligent decision-making and control system that integrates AI with multi-source heterogeneous data. By combining geological exploration, mining process, and real-time monitoring data, and using the D-S theory and CNN-LSTM-Attention network, it enables dynamic identification of subsidence risks and intelligent recommendation of control strategies. Experimental results show that this technology outperforms traditional empirical decision-making methods.