论文部分内容阅读
通过引入全局损失函数,提出了一种全局优化的随机森林模型算法,称为θ-β型随机森林,并且利用改进后的模型对城市遥感图进行了检测与识别,识别准确率与识别速率都得到了一定的提高.方法在经典随机森林模型的基础上加入前向反馈模型(Forward Stagewise Additive Model),通过每一层节点的训练结果干预下一层的训练数据(从而改变阈值θ的选择)与训练步长(β),使得最后训练得到的型随机森林收敛速度更快,预测结果更为准确.
By introducing the global loss function, a globally optimized stochastic forest model algorithm called θ-β stochastic forest is proposed. The improved remote sensing image is detected and identified by using the improved model. Both recognition accuracy and recognition rate are The method is based on the classical random forest model with the help of the Forward Stagewise Additive Model, which intervenes with the training data of the next layer through the training results of each layer of nodes (thus changing the choice of threshold θ) And training step (β), the convergence rate of the final trained random forest is faster and the prediction result is more accurate.