利用并行惯性权重OOL-FA的大数据分类  

Inertia weight OOL-FA for big data classification

作　　者：钟章生[1] 陈世炉 陈志龙[3] ZHONG Zhang-sheng;CHEN Shi-lu;CHEN Zhi-long(College of Computer Information Engineering,Nanchang Institute of Technology,Nanchang 330013,China;6354 Institute,China Shipping Corporation,Jiujiang 332000,China;College of Electronics and Information,Nanchang Institute of Technology,Nanchang 330013,China)

机构地区：[1]南昌理工学院计算机信息工程学院,江西南昌330013 [2]中国船舶总公司第六三五四研究所,江西九江332000 [3]南昌理工学院电子与信息学院,江西南昌330013

出　　处：《计算机工程与设计》2020年第10期2818-2824,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61663033);江西省教育厅科学技术研究基金项目(GJJ180989)。

摘　　要：针对现有大数据分类过程中特征选择算法精度较低,影响后续数据分类算法精度的问题,提出基于惯性权重正交反向学习(OOL)-萤火虫算法(FA)的大数据特征选择算法。借助FA的全局搜索能力,以及OOL分别在收敛速度、收敛精度方面的改进能力,实现数据特征的快速、精确选择,采用结构感知卷积神经网络对大数据特征进行精确分类。在包含6600万个样本和2000个属性的大数据集上进行实验,实验结果表明,所提算法在分类准确率上具有明显的优势。In view of the low accuracy of feature selection algorithm in the existing big data classification process,which affects the accuracy of the subsequent data classification algorithm,a big data feature selection algorithm based on the inertial weight orthogonal reverse learning(OOL)-firefly algorithm(FA)was proposed.With the help of the global search ability of FA and the improvement ability of OOL in convergence speed and accuracy,the data feature was selected quickly and accurately.The structure perception convolution neural network was used to classify the features of big data accurately.Experiments were carried out on a large dataset with 66 million samples and 2000 attributes.The results show that the proposed algorithm has obvious advantages in classification accuracy.

关 键 词：大数据分类 惯性权重OOL-FA算法 结构感知神经网络 Spark框架 特征选择 数据分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]