农用地集约利用评价的新模型研究

The Application of a New Model in Intensive Use Evaluation of Agricultural Land

作为农业大国,中国面临着农用地利用效率低下,耕地数量锐减的现状。如何缓解人地关系、提高农用地集约利用效率是目前中国解决农用地问题的迫切需要。因此,开展农用地集约利用评价具有十分重大的意义。论文以陕西省为例,提出基于动态模糊神经网络的农用地集约利用评价模型,以克服传统方法学习过程慢,易造成规则灾难、过度拟合等问题。为了提高评价的准确性,采用定性与定量相结合的方法挑选评价指标体系,剔除冗余程度高的评价指标:耕地平衡指数和单位面积劳动力人数。对动态模糊神经网络模型进行训练和测试,模型收敛结果良好,误差均不超过3×10-16。为了便于分析评价结果,采用K-means方法将评价分值聚为四类,并与陕西省农用地集约度的空间分异(陕北、关中、陕南)实际情况对比。结果表明,采用新模型的评价结果与陕西省各地级市的实际情况相符,杨凌示范区的集约程度最高。最后,通过逐步回归分析得出,农用地集约度与农民人均纯收入呈显著正相关,其相关系数为0.74,高于人均GDP、城市化水平,农民人均纯收入是影响农用地集约度的主要因素。

As a large agricultural nation, China is faced with low efficiency of agricultural land use and sharp reduction of the quantity of agricultural land. It is urgent to alleviate the relationship between population and land, and improve the efficiency of agricultural land intensive use when China embarks on solving problems of agricultural land. Therefore, it reveals great significance to evaluate the intensive use of agricultural land. Taking Shaanxi Province as an example, the dy namic fuzzy neural network is applied to evaluate the intensive use of agricultural land in order to overcome the low learning process and rule disasters existed in traditional methods. Aiming to im prove the accuracy of evaluation, combination of qualitative and quantitative analysis is used to select evaluation index system without high level of redundancy through eliminating cropland bal ancing index and labor force per hm2 quantitatively and the model receives fine convergence with out exceeding 3 ~ 1016 in errors. In order to analyze evaluation results, evaluation scores ob tained from the new model are clustered into four classes using the Kmeans method. Compared with the actual situation of agricultural land intensive using, the results reveal that intensive de grees of agricultural land distribute with spatial division in accordance with actual situations in Shaanxi and Yangling is the highest in intensive degrees of agricultural land. Finally, intensive degrees of agricultural land and per capita net income of farmers have positive significance correlation through stepwise regression analysis. The correlation coefficient is 0. 74, and is higher than intensive degrees and GDP per capita and urbanization degrees, coming to a conclusion that per capita net income of farmers is the primary factor of affecting intensive degrees of agricultural land.