Abstract: To optimize the hardware deployment of Radio Frequency Identification (RFID) system and improve the deployment efficiency,a RFID system identification rate prediction model based on Neighborhood Rough Set (NRS) and Support Vector Machines (SVM) theory was proposed.The initial influencing factors of RFID system identification rate were reduced by using the neighborhood rough set theory with the principle of minimal correlation and maximal dependency,and the kernel factor subset was obtained.The prediction model of support vector machine was established based on the kernel factor subset,and the prediction model of dynamic RFID system identification rate was constructed by using cross-validation and grid-search adaptive optimization model parameters.The prediction model was tested on the RFID experimental platform.The results showed that the prediction accuracy of the model could reach 92.89%,and the root mean square error value was 0.36.Compared with KNN-Naive Bayesian and other prediction models,the prediction time was shorter and the calculation speed was faster.The validity of the proposed model was verified by an application example of intelligent book management platform.

Key words: radio frequency identification, neighborhood rough set, support vector machines, parameter optimization, prediction

摘要： 为了优化射频识别系统硬件部署,提高部署效率,提出一种基于邻域粗糙集与支持向量机理论的射频识别系统识别率预测模型。首先利用邻域粗糙集理论对影响射频识别系统识别率的初始因子以最小相关和最大依赖度为准则进行属性约简,筛选出核因子集。基于该核因子集建立了支持向量机预测模型,并使用交叉验证与网格搜索法自适应寻优模型参数,构造动态射频识别系统识别率预测模型,并对射频识别实验平台进行测试。结果表明,该模型预测分类准确率可达92.89%,均方根误差值为0.36,相比较基于K最近邻—朴素贝叶斯等其他分类预测模型,预测时间更短,运算速度更快。最后,通过智慧图书管理平台的应用实例,验证了所提模型的有效性。

关键词: 射频识别, 邻域粗糙集理论, 支持向量机, 参数优化, 预测