无线传感网络监测数据分类算法仿真分析  

Simulation Analysis of the Monitoring Data Classification Algorithm for Wireless Sensor Networks

作　　者：洪月华[1] 

机构地区：[1]广西经济管理干部学院计算机系,广西南宁530007

出　　处：《微电子学与计算机》2013年第10期149-152,共4页Microelectronics & Computer

基　　金：广西自然科学基金青年项目(2012GXNSFBA053178);广西混杂计算与集成电路设计分析重点实验室开放基金课题(2012HCIC05)

摘　　要：研究无线传感网络数据的准确分类问题.针对无线传感网络监测得到的数据属性呈现高冗余特征,用传统的BP神经网络进行分类易陷入局部最优解、泛化能力差、收敛速度慢与精度低等问题,造成数据很难被准确分类,提出用免疫算法优化BP神经网络的分类算法.该算法利用免疫算法所具有的全局收敛特性和个体多样性保持机制,全局搜索优化BP神经网络的权值,再用BP算法对其开展局部搜索工作.仿真实验结果证明,该算法能够有效克服训练经典BP神经网络时容易陷入局部极值的不足,加快了网络的收敛速度,传感器网络数据分类识别的准确率也得到了大幅度提高.The problem of data classification for wireless sensor network is researched. The data monitored by wireless sensor network is high redundant, so if traditional BP neural network is used for classification will result in local optimal solution, poor generalization ability, slow convergence speed and low accuracy problem. Therefore data is difficult to be accurately classified. In order to solve the problem immune algorithm is used to optimize BP neural network classification algorithm. Based on the global convergence and individual diversity maintaining mechanism of immune algorithm, the algorithm is used to search globally and optimize BP neural network weights, and then carry out local search with BP algorithm. The simulation results show that, the algorithm can effectively overcome the shortcoming that training classic BP neural network can easily fall into local extreme, so it can accelerate the network convergence speed and the accurate rate of sensor network data classification has been greatly improved.

关 键 词：免疫算法 BP神经网络 无线传感器网络 数据挖掘 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]