Internet of things traffic classification based on lightweight random forest algorithm
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
维普
万方数据
为解决物联网设备资源受限、平衡流量检测精度与时间开销等问题,提出一种FastSplit-RF(random forest with fast split)的轻量化分类算法.针对物联网流量设计一个通用的特征提取流程,在随机森林算法基础上,使用多臂赌博机策略代替节点分裂的遍历过程,实现对节点的快速分割,完成高效、轻量化的物联网流量分类.实验验证,FastSplit-RF相较随机森林算法,在准确率提升了2.45％的同时,检测速度增快了62.16％,内存占用减小了48.68％.
To address challenges such as limited resources in IoT devices and prolonged delays in traffic detection and classifica-tion,a lightweight classification algorithm named FastSplit-RF(random forest with fast split)was introduced.A universal fea-ture extraction process for IoT traffic was designed,and on the basis of the random forest algorithm,a multi-armed bandit stra-tegy was employed to replace the traversal process during node splitting.Rapid segmentation of nodes was enabled,allowing efficient and lightweight IoT traffic classification.Experimental results indicate that,compared to the random forest algorithm,a 2.45％improvement in accuracy is achieved using FastSplit-RF,accompanied by a 62.16％increase in detection speed and a 48.68％reduction in memory usage.
维普
万方数据
