摘要
为建立一种塑料吸管物证的高效、准确分类方法,利用红外光谱法对来自全国的4个品牌共42个塑料吸管样本进行了检验。经过前期光谱预处理后,利用主成分分析法提取出了25个主成分,累积方差贡献率为99. 689%,并将其作为判别变量进行判别分析。判别结果区分效果良好但交叉验证正确率仅为73. 8%,有待进一步提高。因此将判别得分作为特征变量导入K值为1的K近邻算法中,构建起了分类正确率为100%的K近邻算法模型,实现了对塑料吸管物证的准确分类。
In order to establish an efficient and accurate classification method for the physical evidence of plastic straws,a total of42 plastic straw samples from 4 brands across the country were tested by infrared spectroscopy. After pre-processing of the original spectrum,25 principal components were extracted by principal component analysis,and the cumulative variance contribution rate was99. 689%,which was used as a discriminant variable for discriminant analysis. The discriminating results are good,however,the accuracy rate of cross-validation is only 73. 8%,which needs to be further improved. Therefore,the discriminant score is introduced into the K-nearest neighbor algorithm with K value of 1 as a feature variable,and the K-nearest neighbor algorithm model with classification accuracy rate of 100%is constructed to achieve accurate classification of the physical evidence of plastic straws.
作者
姜红
马枭
杜岩
JIANG Hong;MA Xiao;DU Yan(School of Forensic Science,People s Public Security University of China,Beijing 100038,China)
出处
《塑料工业》
CAS
CSCD
北大核心
2020年第5期112-116,共5页
China Plastics Industry
基金
中国人民公安大学2019年度基本科研业务费重点项目(2019JKF222)
国家重点研发计划项目(2017YFC0822004)。
关键词
红外光谱法
判别分析
K近邻算法
塑料吸管
Infrared Spectroscopy
Discriminant Analysis
K-nearest Neighbor Algorithm
Plastic Straws
