Abstract: Accurately predicting the reclamation potential of rural settlements is critical for optimizing regional land use patterns. However, many existing approaches lack reliable verification of their predictions. This study proposes a potential prediction model for decision-making on rural settlement reclamation, using Gradient Boosting Regression Tree (GBRT) and spatiotemporal land-use data. Because the reclamation potential of rural settlements is influenced by a complex interplay of physical and socio-economic factors. A systematic analysis was conducted to identify the influencing factors on the release of land reclamation potential in rural areas, integrating them into a system of 35 factors, including topography, location, urban-rural development, government financial pressure, agricultural development, food security, and distribution characteristics of rural settlements and cropland. The spatiotemporal land use data were selected to create the training and test samples for the model. The GBRT also presented the high accuracy, efficiency, and resistance to overfitting. The improved model was used to automatically identify the potential release characteristics of rural settlements under the comprehensive influence of multiple socio-economic factors. A case study was conducted to evaluate the effectiveness and performance of the improved model in the Hunan Province of the South-Central China. The data used in the case study was collected from the 102 county-level units in Hunan, where 70% for training and 30% for validation. The cross-validation experiments show that the better performance was achieved in the improved model, with the determination coefficient of R2=0.976 5 and an average absolute percentage error of 11.64%. The relative error was about 6.6% between the predicted and the actual potential between 2009 and 2019. Therefore, the prediction accuracy of the improved model was fully met the requirements of the planning decisions. According to the prediction, the total reclamation potential of rural settlements from 2020 to 2035 in Hunan was approximately 36 050.26 hm2, equivalent to 4.58% of the current rural construction land area in 2020. The predicted potential characteristics and distribution were in line with the actual reclamation potential from 2009 to 2018, indicating the feasibility of the predicted model. The areas with the greatest potential were primarily located in the "Changsha-Zhuzhou-Xiangtan" area, while the areas around Dongting Lake and the central Hunan Basin presented relatively fast economic development and urbanization, indicating a large contradiction between the demand and supply of land for construction. Therefore, the rural settlements to be reclaimed were in plain areas with convenient transportation and low difficulty of reclamation, where contiguous operation with surrounding arable land after reclamation could yield better economic benefits. These regions can serve as the major area for comprehensive land consolidation from 2020 to 2035. The improved model can provide accurate and reliable guidance for comprehensive land consolidation and land use planning, benefiting provincial and local governments. In conclusion, the accurate prediction achieved in the reclamation potential of rural settlements will help optimize land use patterns and promote the sustainable development of the region.