中文摘要：

建设用地是城市发展的重要因素,对建设用地规模的预测可以为土地利用总体规划提供参考数据和技术支持。笔者以连云港市为例,收集了2004—2013年有关建设用地规模的社会经济统计数据,采用主成分-BP神经网络模型对连云港市2014—2020年建设用地规模进行预测,得出7年连云港市建设用地规模的预测结果。研究得出主要结论：（1）主成分分析结果显示社会经济的发展、人口和基础设施的变化以及环境的改善从不同方面影响着建设用地的规模;（2）笔者构建的BP神经网络模型误差率较低、拟合效果较好且对于训练集以外的新样本数据具有较好的泛化能力,说明所建模型具有可靠性,可以进行预测;（3）连云港市2014—2020年的建设用地规模呈现逐年扩张的趋势,年均增长率为0.97%,连云港市应采取有效措施控制建设用地规模并且合理保护耕地,使得建设用地面积的增长控制在合理的范围之内。主成分-BP神经网络模型不仅能够对影响建设用地规模的因素进行全面分析,同时可以得到精度较高的建设用地规模预测数据,因此能够较好地应用于建设用地规模预测。

英文摘要：

Construction land is an important factor in the process of urban development,prediction of construction land scale can provide reference data and technical support for general land use planning.The authors collected sociol and economic data from 2004 to 2013 which influenced the construction land scale of Lianyungang,and applied the principal component-BP neural network model to predicting the scale of construction land in Lianyungang from 2014 to 2020,and obtained seven years’prediction results of construction land scale in Lianyungang.The results showed that（1）the principal component analysis indicated that the development of social economy,the changes of population and infrastructure,the improvement of the environment mainly influenced the construction land scale in Lianyungang from different aspects;（2）the error rate of the BP neural network model built in this research was low and its imitative effect was good,in addition,the model also had good generalization ability to the new sample data from outside of the training sets,which illustrated that the model was credible,and it could be applied to predicting the construction land in the future;（3）the construction land scale of Lianyungang showed an expansion trend from 2014 to 2020,the average annual growth rate was 0.97%,effective measures should be taken to control the construction land scale and protect arable land,making the growth of construction land area controlled in a reasonable range.Principal component-BP neural network model could not only analyse the influencing factors of the construction land scale,but also get high precision prediction data of construction land scale,as a result,this model should be well applied to construction land scale prediction.