Abstract: Current research on the developing heights of water-conducting fracture zones mainly focuses on the heights of water-conducting fracture zones in single coal seam mining, while research on the prediction of the developing heights of water-conducting fracture zones in extremely thick coal seams mining is relatively scarce. Based on the geological conditions of the extremely thick coal seams in the Jurassic coalfields of Xinjiang, this research selected the parameters of the typical working face 9-15 (08) in the Sulphur Gully Coal Mine of the Zhunan Coalfield in Xinjiang, quantitatively evaluated the development characteristics and evolution patterns of the overlying rock fracture fields under layered full-mechanized mining of extremely thick coal seams through numerical simulations and fractal geometry theory analysis. A prediction model was developed for the heights of water-conducting fracture zones in layered mining of extremely thick coal seams based on Particle Swarm Optimization Support Vector Machine Regression (PSO-SVR). The research results showed that: ① During layered mining of extremely thick coal seams, the hard rock strata and inferior key strata within the capping range exhibited a hinged structure, and the overall deformation and failure range of the overlying rocks presented an arched structure. ② The impact of mining activities caused the roof overlying rocks to fracture and collapse, with horizontal fractures continuously developing and vertical fractures extending upwards. The water-conducting fracture zones rose continuously, and the fractal dimension increased rapidly. As the working face continued to advance, the horizontal fractures within the overlying rock layers were compacted by the layer above, the fracture aperture decreased, and the fractal dimension gradually reduced. ③ During layered mining, the fractal dimension of fractures generally exhibited four stages: ascending dimension stage, dimension reduction stage, stationary stage, and fluctuating stage. ④ The PSO-SVR model was evaluated using indicators including mean absolute error (MAE), mean bias error (MBE), and correlation index R². The model showed that correlation index R²＞0.90, MAE＜6.5 m, and MBE within ±0.5 meters, indicating that the PSO-SVR model was capable of predicting the heights of water-conducting fracture zones in layered full-mechanized mining. ⑤ By substituting the data from the working face 9-15(08) into the PSO-SVR model, the absolute error between the predicted and observed values was 12.52 m, and the relative error was 4.86%, indicating that the PSO-SVR model could effectively and accurately predict the heights of the water-conducting fracture zones in extremely thick coal seams mining.