An evaluation method of wasteland landscape restoration plan based on BP neural network model

摘要

Wasteland Landscape Restoration has become an important method of wasteland control. Scientific methods are crucial to the selection of appropriate Wasteland Landscape Restoration Plan. This paper builds a three-hierarchy BP neural network model and selects six indicators including spatial layout, functional localization, pollution control, resources utilization, cultural features and road transportation for the evaluation and prediction of Wasteland Landscape Restoration Plans. The results indicate that BP neural network model is able to fit the original data accurately, and high-precision prediction is achieved for most samples. The relative error between predicated value and true value is less than 10%. It is suggested that BP network, after effective training, can be applied in the evaluation of Wasteland Landscape Restoration Plan, with a high precision of prediction and good generalization ability. This method does not require the construction of complicated parametric equations and has a strong self-learning ability, simple and practical structure as well as bright application prospect.