生态环境学报2018，Vol.27Issue(5)：964-973,10.DOI:10.16258/j.cnki.1674-5906.2018.05.023
基于GA-BP神经网络的珠三角耕地质量评价
Evaluation of Cultivated Land Quality in Pearl River Delta Based on GA-BP Neural Network
摘要
Abstract
The cultivated land and its quality is of vital importance to the security of grain, the benign development of economy, and the harmony and stability of the society. Under the trend of high-quality cultivated land resources being continuously invaded and declining in numbers, it is a reliable choice to actively carry out the research of cultivated land quality so as to adapt to the social and economic development. Scientific evaluation of cultivated land quality can accurately understand the present situation of quality and the spatial distribution characteristics, which is of great significance in guiding the rational utilization and protection of limited cultivated land resources, as well as in realizing the comprehensive balance and management of the quantity-quality of cultivated land. Constructing reasonable evaluation index systems and exploring effective evaluation methods have become important subjects of the current research on the quality evaluation of cultivated land. Although there are many researches on the comprehensive evaluation of cultivated land quality, they are more or less subjective or defective. The purpose of this paper is to explore an intelligent evaluation method of cultivated land quality, avoid setting index weights, and improve evaluation efficiency. Taking the cultivated land in the Pearl River Delta Region as the research subject, this paper constructed an index system of comprehensive evaluation of cultivated land quality, which was suitable for the area on the aspects of nature quality, economy quality, utilization quality and ecology quality. On the basis of BP (Back Propagation) neural network model, a genetic algorithm was introduced to design an evaluation method based on Genetic Algorithm- Back Propagation (GA-BP) neural network. In the Pearl River Delta, 4 000 representative samples were selected, of which 3 000 were used as training samples, 500 were used as test samples and 500 were used as test samples. The model was trained by GA-BP neural network model, and the results of cultivated land quality evaluation were output by simulation, so as to analyze the distribution of cultivated land quality grade. The experiment results showed that the training times of GA-BP neural network model were obviously less than that of BP neural network, and the difference between the maximum and minimum mean square error (MSE) is 0.105 1 less than that of BP network model, which was closer to the actual cultivated land quality grade. Hence it provided better stability and fitness for the quality evaluation of cultivated land. The results of cultivated land quality evaluation showed that the quality of cultivated land in the Pearl River Delta region was generally good, in which the proportion of two or three places was the largest, accounting for 52.94% of the total cultivated land area. The cultivated land quality was basically in accordance with the normal distribution trend, showing an obvious geographical distribution pattern, generally high in the middle and low in the periphery. There was a huge difference in the distribution of cultivated land quality grades in each area. This paper enriches and improves the evaluation index system and evaluation method of cultivated land quality in a wide range of areas, provides a basis for the sustainable use of cultivated land resources in the Pearl River Delta, and offers some reference for other similar researches.关键词
耕地/质量评价/GA-BP神经网络/珠三角Key words
cultivated land/quality evaluation/GA-BP neural network model/Pearl River Delta Region分类
资源环境引用本文复制引用
叶云,赵小娟,胡月明..基于GA-BP神经网络的珠三角耕地质量评价[J].生态环境学报,2018,27(5):964-973,10.基金项目
广东省科技计划项目(2013A040600002) （2013A040600002）
广东省科技发展专项资金项目(2017A020208056) （2017A020208056）
