基于自适应指数蝙蝠和SAE的并行大数据分类

Parallel Big Data Classification Method Based on Adaptive Exponential Bat and Stacked Autoencoder

为解决深度学习进行大数据分类时效率低的问题,本文提出一种基于自适应指数蝙蝠和堆叠自编码器(SAE)的并行大数据分类方法.在并行计算框架中,Map阶段使用自适应指数蝙蝠算法进行特征选择,自适应指数加权移动平均值蝙蝠算法(AEB)由指数加权移动平均值(EWMA)和自适应权重策略得到.将选择的特征作为Reduce输入进行大数据分类,Reduce阶段使用AEB算法训练的深度堆叠自动编码器(SAE)进行分类,进一步提高了分类精度.实验结果表明,针对不同的训练数据百分比,本文所提方法在准确度和真正例率(TPR)性能方面优于其他现有方法.

A parallel big data classification method based on adaptive exponential bats and SAE has been proposed to solve the problem of low efficiency when classifying big data by deep learning.In the parallel computing framework,AEB algorithm is used to select features in the Map stage.AEB is obtained according to the exponential weighted moving average(EWMA)and adaptive weight strategy.Then the selected features are used as the input of Reduce for big data classification.In the Reduce stage,the deep stacked autoencoders trained by AEB algorithm is used for classification,which further improves the classification accuracy.The experimental results show that the proposed method is superior to other existing methods in terms of accuracy and TPR performance for different percentage of training data.