<p>Fully understanding the effects of dry–wet cycles on the micro-macroscopic structure and mechanics of coal is of great significance to the monitoring and prevention of kinetic disasters in coal mines. The pore structure of coal samples with different dry–wet cycles (0, 3, 6, 9) was comprehensively characterized using microscopic probing means. Then, uniaxial compression tests were conducted on the coal samples under the same loading conditions, and the loading process was monitored by a full-information acoustic emission system to reveal the macroscopic mechanical performance of coal under the influence of dry–wet cycles. The results showed that the cumulative porosity of the coal samples was about 99.14% higher after 9 times of dry–wet cycling; the peak strength of the coal gradually decreased with the increase of cycling times. With the increase of the number of dry–wet cycles, the damage degree and damage range of the coal samples were larger and larger; with the increase of porosity, the <i>b</i> value decreased faster, the proportion of tensile cracks increased gradually, and the energy released by the damage was more concentrated and high intensity. Finally, the stress intensity of coal samples was predicted based on machine learning, and the reliability of different models for classification of damage stages was compared. The innovative construction of LSTM–Informer–XGBoost intelligent prediction system realizes the accurate classification of coal body damage stages, breaking the bottleneck of traditional monitoring methods in identifying the damage accumulation process. The research results provide a multi-scale theoretical framework of pore evolution–mechanical response–intelligent prediction for the early warning of coal mine dynamic disasters, and provide a certain theoretical guidance for the construction of a disaster prevention and control system for water intrusion in deep coal seams.</p>