基于Boltzmann机的物联网数据分类方法研究  

Research of Data Classification Method in Internet of Things Based on Boltzmann Machine

作　　者：刘健[1] 常莲[2] 

机构地区：[1]金肯职业技术学院,南京210000 [2]南京航空航天大学金城学院,南京211156

出　　处：《激光杂志》2016年第10期116-120,共5页Laser Journal

基　　金：国家自然科学基金(编号:61170121)

摘　　要：物联网是传统互联网在现代社会应用的拓展和延伸。本文通过分析物联网系统进行数据分类的必要性,同时根据Boltzmann机的应用特性,提出了基于Boltzmann机的物联网数据分类方法。通过对Boltzmann机进行网络建模,制定运行规则和自联想记忆学习规则,完成对物联网数据样本的学习及训练过程,实现物联网数据分类。通过在Matlab中完成学习样本的构建和分类方法的具体实现,对本文提出的数据分类方法进行测试、验证和分析。递减的能量函数曲线和交叉熵曲线证明系统可以达到稳定,且最终分类结果的分布概率与期望分布结果基本吻合。高准确度的分类方法能够满足实际应用的需要,也进一步验证了本文所提方法的可行性。总之,基于Boltzmann机的物联网数据分类方法效果突出。The Intemet of things is the development and extension of traditional Internet applications in modern society, at the same time, along with the advent of the era of big data. This data classification method in internet of things based on Boltzmann machine is proposed through the analyzing the necessity of the IoT system for data classification and Boltzmann machine at the same time. Based on the network modeling of Boltzmann machine and formulating operation rules and the associative memory learning rules, the data classification of IoT is realized after completing the IoT data samples＇ learning and the training process. Through building learning samples in the Matlab and concrete re- alization of the method, the proposed data classification method is test, verified and analyzed. Decreasing energy func- tion curve and cross entropy curve prove the stability of the system, and the distribution of the final classification result is be identical with the expected results. High accuracy of classification method can meet the needs of practical appli- cation, and further verifies the feasibility of the proposed method in this paper. All in all, data classification method in internet of things based on Boltzmann machine has prospecting effect.

关 键 词：物联网 BOLTZMANN机 数据分类 自联想记忆 概率分布 

分 类 号：TN919[电子电信—通信与信息系统]