The focus of China’s coal mining in recent years has shifted to western China, which has a fragile ecological environment. The Jurassic coal in this area is buried at shallow depths and is very thick. Currently, the two main issues are the inrush of underground water and the disruption to shallow surface water supplies brought on by mining. Accurately predicting the height of the water-conducting fractured zone (WCFZ) is the key to preventing roof water damage and attaining water conservation in this mining area. A nearest neighbor bootstrap regression (NNBR) model was established to predict the height of the WCFZ for working face 117 of the Jinjitan coal mine in western China, which uses fully mechanized top coal caving longwall mining. The model was verified based on the measured heights of the WCFZ (30 groups of measured data) using the actual mining thickness (> 8 m) and working face width as predictive factors. The mean error of the model’s prediction was 1.5 m and the mean absolute percentage error was 13.6%, which was better than the result predicted using the established conventional empirical formula. The model was improved by applying a safety factor of 1.3, which met the 96.7% reliability requirements. The height of the WCFZ of working face 117 was predicted using this model, and the predicted value was compared with values measured by optical fiber monitoring and simple hydrological observations. The height of the WCFZ measured by optical fiber was 184.8 m, the predicted value was 196.0 m, and the relative error of that prediction was 6.0%. The hydrologically observed value was 221.0 m, the predicted value for that location was 225.7 m, and the relative error was 2.1%. Thus, the NNBR model is capable of reliably predicting the height of the WCFZ and it appears that it can be used to compute the safe water control height above mechanized top coal caving mines in western China’s very thick Jurassic coal seams.