| Scientifically revealing the driving mechanism of industrial water use change is the basis for accurately predicting industrial water demand. To address the current issue that the driving factors of industrial water use change and their action mechanisms remain unclear, this study takes Baotou City, Ordos City, and Yulin City—energy-rich regions in the Yellow River Basin—as research objects. It adopts big data mining to construct a knowledge graph for driving industrial water use change by analyzing the association network between industrial indicators and water use. Additionally, it couples the Particle Swarm Optimization-Support Vector Machine (PSO-SVM) main control factor analysis with the symbolic regression algorithm to establish a non-parametric deep learning model, fits the constitutive equation of industrial water use evolution, and reveals the driving mechanism of industrial water use evolution. A set of water demand prediction technologies driven by both big data mining and physical mechanisms is further developed, with which the industrial water demand from 2025 to 2035 is predicted. The results show that due to differences in industrial structure, resource endowments, and water resource conditions, the evolution mechanisms of industrial water use in different regions exhibit both industrial similarity and regional difference. Industrial added value is identified as a common driving indicator, with a contribution rate of over 0.06 in all regions. Water resource development and utilization rate is a common inhibiting indicator, with a contribution rate of below -0.06 in all regions. Under the combined action of multiple driving indicators, there are significant differences in the changing trends of future industrial water demand among Baotou, Ordos, and Yulin. Specifically, the industrial water demand of Baotou City will decrease by 12.2%, while that of Ordos City and Yulin City will increase by 10.5% and 11.6% respectively. |