基于小样本学习的物联网异常状态修正算法  

Abnormal State Correction Algorithm of Internet of Things Based on Small Sample Learning

作　　者：燕敏 阮秀琴 赵阳[2] 郑宏涛 YAN Min;RUAN Xiu-qin;ZHAO Yang;ZHENG Hong-tao(Information Center,Xi'an Shiyou University,Xi'an Shanxi 710065,China;Information Management Office,Xi'an University of Technology,Shanxi Xi'an 710048,China)

机构地区：[1]西安石油大学信息中心,陕西西安710065 [2]西安理工大学信息化管理处,陕西西安710048

出　　处：《计算机仿真》2022年第8期389-393,共5页Computer Simulation

摘　　要：传统物联网异常状态修正算法需大量状态样本,导致算法耗时较长,精度偏大。为解决上述问题,提出基于小样本学习的物联网异常状态修正算法。获取物联网状态的空间相关性特征,采用欧式度量方法完成小样本学习。搭建小样本学习网络,得到物联网异常状态检测结果,构建RBF神经网络修正模型。实验结果表明:与两种传统算法相比,所提算法物联网异常状态修正正确率平均值分别提高了20.02%与24.87%,算法时延平均值分别降低了1.33s与1.48s,实验所得数据充分验证了提出算法的物联网异常状态修正效果更好。The traditional abnormal state correction algorithm of the Internet of things requires a large number of state samples,resulting in a long time-consuming algorithm and high accuracy.In order to solve the above problems,an abnormal state correction algorithm of Internet of things based on small sample learning is proposed.The spatial correlation characteristics of the state of the Internet of things were obtained.Based on the Euclidean measurement method,small sample learning was achieved.Small sample learning network was built to get the abnormal state detection results of the Internet of things,and to build RBF neural network correction model.The results show that the algorithm has excellent effect in correcting the abnormal state of the Internet of things.Compared with the two traditional algorithms,the average accuracy of the algorithm is increased by 20.02%and 24.87%respectively,and the average delay is reduced by 1.33s and 1.48s respectively.

关 键 词：小样本学习 物联网 状态修正 神经网络 模型参数 

分 类 号：TN919.5[电子电信—通信与信息系统]