The problems of machine learning in shallow for prediction were concerned because the learn-ing and training samples was too large and complex function fitting ability was weak. A cultivated land acreage prediction based on deep belief support vector regression algorithm was proposed. Firstly, an deep belief support vector regression model was constructed by a RBM with Gaussian distribution function layer nodes, amiddle layer RBM and a support vector regression. Secondly, suitable and accessible train-ing data was used to get the parameters of the prediction model. Finally, comparsion between the deep belief support vector regression land acreage prediction and other typical farmland prediction algorithm were conducted. Results showed that the propose method was workable and effective, which perform bet-ter than the typical farmland prediction algorithm with the same data and condition.%针对目前浅层机器学习预测方法所需学习和训练的样本过大及拟合复杂数据能力弱等不足,提出一种基于深度学习思想的深度信任支持向量回归( support vector regression,SVR)的耕地面积预测方法。首先,搭建由1层高斯分布函数显层节点的RBM、多层隐层RBM和1层支持向量回归机构成的深度信任支持向量回归预测模型；其次,选取较为合适和易得的训练数据,通过样本训练和测试确定预测模型的具体结构参数；最后,通过实验将深度信任支持向量回归耕地面积预测方法与其他典型的耕地面积预测算法相比较。结果表明,提出的耕地面积预测方法可行、有效,在相同的数据和平台下,其预测精度高于其他具有代表性的耕地面积预测算法。
展开▼
