摘要
传统设计模型构建方法存在校正集和验证集代表性弱的问题,导致内在质量检测误差大,提出基于大数据挖掘技术的卷烟内在质量自动检测模型构建。通过对大数据挖掘平台的架构,可快速对卷烟数据进行采集与分析;利用大数据挖掘技术对卷烟感官质量进行挖掘,并聚类分析,区分不同部位间内在质量的化学成分差异;研究卷烟属性、感官质量关联性、内在质量的物理属性和化学属性,构建自动检测模型。实验验证结果可知,该模型的内在密度校正集与验证集相差最小值为1. 9,相差最大值为9. 4,远小于传统模型的最大相差值和最小相差值,当主成分维数为9个时,该模型交叉验证均方根最小,且内在质量检测较为精准,说明大数据挖掘技术下,所构建得内在质量模型检测效果良好。
The traditional design model construction method has the problem of the weak representation of the correction set and the validation set, which leads to the large error in the internal quality detection. Based on the big data mining technology, the automatic detection model of cigarette internal quality is built. Through the platform of large data mining, fast acquisition and analysis of data mining technology on cigarette ; cigarette sensory quality and the use of big data mining, clustering analysis, distinguish the difference of chemical composition of the intrinsic quality of different parts ; physical and chemical properties of tobacco sensory quality attributes, relevance, quality, construction of automatic detection model. The experimental results show that the model of the internal density calibration set and validation set is the minimum value is 1.9, the maximum value is 9.4, the maximum value and the minimum value difference is far less than the traditional model,when the principal component dimension is 9,the model of cross validation RMS minimum, and internal quality test is more accurate, the data mining technology, establishes internal quality model of good detection effect.
作者
凌军
杨乾栩
张天栋
张玲
唐军
杨建云
LING Jun;YANG Qianxu;ZHANG Tiandong;ZHANG Ling;TANG Jun;YANG Jianyun(Yunnan medium Tobacco Industry Co.,Ltd.,Kunming,650231)
出处
《自动化与仪器仪表》
2018年第11期178-182,共5页
Automation & Instrumentation
基金
云南中烟工业有限责任公司重点项目大数据优化提升云产卷烟内在质量的研究(2015CP02)
关键词
大数据挖掘
卷烟
内在质量
自动检测
模型构建
Large data mining
cigarette
internal quality
automatic detection
model construction
