收敛性随机森林模型及其遥感应用

Convergent Random Forests Model and Its Application in Remote Sensing

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.

A improved Random Forests called θ-β Random Forests（θ-β RFs） is proposed in this paper.And the detection and identification of the remote sensing images of cities are done using θ-β RFs.The accuracy and speed rate of the experiment result is proved to be better.This model combines RFs with Forward Stagewise Additive Model,alternating the results of the next nodes by the results of the current nodes.θ-β RFs has a greater convergence rate and more accurate results.