[Objective] The paper aims to compare the accurracy of BP neutral network and GM(1,n) in predicting construction land changes, which is beneficial to understand the urban development and provide refernces for general land planning. [Methods] With Siping City as the research object, we selected impact factors with the perspective of “city-rural integration” and used factor analysis to estimate the influence of construction land expansion and choose indicators. We then simulated and compared the predictions of construction land in 2012, 2013 and 2014 in Siping City using the BP neural network and grey model. [Results] The relative error with BP neural network was 0.8%, 1.1% and 2%, and the gray GM(1.1) model was 0.04%, 0% and 3.2% respectively. The BP neural network are better than GM(1.1) model. [Conclusion] BP neural network can provide a higher accuracy.