基于支持向量机的北京市湿地变化预测研究

为了实现北京市湿地的可持续发展与科学管理，利用遥感和GIS技术对北京区域遥感数据进行分析，进而获取湿地不同类型数据，由于湿地面积是典型的小样本数据，将支持向量机引入到时间序列模型定阶的方法中，然后采用K交叉验证方法寻找最优参数，建立北京湿地变化预测模型。通过对北京湿地历年数据进行模拟，并与RBF神经网络的预测模型作比较来验证SVM预测模型的有效性，运用此模型预测北京湿地未来的变化。结果表明：时间序列模型预测湿地变化有较高的预测精度和较强的泛化能力。预测结果显示：北京湿地在未来几年内，水库、河流，运河，沟渠面积将逐年减少，水稻田面积将持续下降，养殖面积将增大。其预测结果符合现实北京湿地变化趋势，研究结果为北京湿地的可持续发展和科学管理提供了依据。

Wetland Change Forecast in Beijing Based on Support Vector Machines

In order to achieve sustainable development and scientific management of Beijing wetland,the author analyzed Beijing remote sensing data for acquiring different types of wetland data,by using remote sensing and GIS technology.Because wetland area is typical of small sample data,support vector machine was introduced into the time series model fixed order method,and then applied K cross validation methods to seek optimum parameters,established the Beijing wetland changes prediction model.Through the historical data of Beijing wetland was simulated,and comparing with RBF neural network,the author verified the validity of the forecast model of SVM,and then used this model to predict Beijing wetland future changes.The forecast results showed that the series model had higher precision of prediction and strong generalization ability,Beijing wetland in the next few years,reservoir,rivers,canals and ditches area would gradually decline,rice paddies area would continue to fall,breeding area would be increased,the predicted results conformed to reality Beijing wetland trend.The findings had provided the basis for the sustainable development and the scientific management of Beijing’s wetlands.