基于改进多核学习的多传感数据分类方法研究

Classification of internet multisensor data based on improved multi-kernel learning

物联网(internet of things,IoT)技术中结合多个数据源互补信息提高数据分类准确率的研究受到了越来越多的关注。针对物联网无线传感器采集到数据的多源异构特性,给出了一种基于改进多核学习支持向量机(improved multi-kernel learning-support vector machine,IMKL-SVM)的IoT数据分类方法。传统的多核学习方法中核函数主要是采用经验法选取核函数类型及参数,本文改进方法在确定核函数类型及参数时分为两步:首先采用交叉验证方法初步确定核函数类型及参数;其次在第一步结果中利用支持向量机(SVM)同时训练样本和优化多核函数的类型及参数。实验中针对温度、湿度、光照、大气压力等4种数据设计了两组数据——第一组数据被标记为上午、中下午、傍晚、夜间4类,第二组数据被标记为白天、傍晚、夜间3类,比较了本文的IMKL-SVM方法、单核SVM方法及传统MKL-SVM方法在两组数据集上的分类准确率。此外,针对UCI公开数据集AReM进行了分类实验,实验结果表明IMKL-SVM方法针对具有多源异构特性的物联网数据实现了较高的分类准确率。

Combining complementary information from multiple data sources to improve the accuracy of data classification has attracted more and more attention in the internet of things(IoT)technology.In order to better classify the multi-source heterogeneous characteristics of data collected by IoT wireless sensors,we have developed an IoT data classification method based on an improved multi-kernel learning support vector machine(IMKL-SVM).Kernel function types and parameters in traditional multi-kernel learning methods are mainly selected according to rule of thumb.Our improved method has two steps for determining types and parameters of the kernel function.The first step is to utilize a cross-validation method to determine types and parameters of the kernel function.The next step is to train the samples and to optimize simultaneously the types and parameters of the multi-kernel function using the SVM.In experiments,two sets of data were designed for four types of data:temperature,humidity,light and atmospheric pressure.One set of data was labeled as morning,mid-afternoon,evening and night,and another set of data was labeled as day,evening and night.We compared the classification accuracy for the two sets of data using our improved method,single-kernel SVM and traditional MKL-SVM.In addition,a classification experiment was conducted on the UCI public dataset AReM.The results showed that our IMKL-SVM method achieves higher classification accuracy for IoT data with multi-source heterogeneous characteristics.