基于深度信念网络与随机森林的药品鉴别方法  

Medicine Identifying Method Based on Deep Belief Network and Random Forests

作　　者：伏为峰 杨辉华 刘振丙[1] 冯艳春 FU Wei - feng;YANG Hui - hua;LIU Zhen - bing;FENG Yan- chun(School of Computer Science and Information Security, Guilin University of Electronic Technology, Guilin Guangxi 541004, China;School of Automation, Beijing University of Posts and Telecommunications, Beijing 100876, China;National Institutes for Food and Drug Control, Beijing 100050, China)

机构地区：[1]桂林电子科技大学计算机与信息安全学院,广西桂林541004 [2]北京邮电大学自动化学院,北京100876 [3]中国食品药品检定研究院,北京100050

出　　处：《计算机仿真》2018年第4期325-330,共6页Computer Simulation

基　　金：国家自然科学基金项目(21365008;61562013);广西重点研发计划项目(桂科AB16380293)

摘　　要：为解决近红外光谱药品鉴别中,光谱数据特征维数较高,传统的浅层特征提取方法学习能力不足的问题,提出了基于深度信念网络和随机森林的药品鉴别算法。算法将深度信念网络与随机森林相结合,利用了深度信念网络对高维特征向量强大的分析和提取能力,以及随机森林良好的分类性能,将深度信念网络作为特征提取器,随机森林方法作为分类器。通过用三种近红外光谱数据集进行实验验证,与其它传统算法作对比,证明了改进方法的有效性,且随着原始输入特征数的提升,算法性能越显优越。To solve the problem that the spectral feature of the near infrared spectrum is too high that the tradi- tional shallow feature extraction method is incapable to extract it, an algorithm based on deep belief network and ran- dom forests is put forward. This algorithm combines the deep belief network and the random forest, and it makes use of the powerful ability to analyze and extract the high dimensional eigenvector with the deep belief network, and the good classification performance of the random forests method. The deep belief network is taken as the feature extractor and the random forests method as the classifier. The experimental results show that the proposed method is effective compared with other traditional algorithms on three data sets, and the performance of the algorithm is improved with the increase of the original input characteristic number.

关 键 词：近红外光谱 深度信念网络 随机森林 药品鉴别 深度学习 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]