摘要
针对卷烟感官评估中存在的代价敏感问题,将基于代价敏感的反馈神经网络应用于卷烟感官评估中。为了验证方法的有效性,结合烟草企业生产实际设置代价矩阵,并利用烟草公司提供的数据进行了对比试验。结果表明,与代价不敏感方法相比,本方法在错分总代价,高代价类别识别率以及平均分类准确率3个方面均有显著改善。
Arming at the cost-sensitive problems in cigarette sensory evaluation, Cost-Sensitive Back-Propagation Neural Networks（CSBPNN） was employed in this paper to deal with the problems derived from cigarette sensory evaluation. In order to verify the effectiveness of our methodology, the cost matrix was obtained based on production practice and the comparative experimental study was carried out by using dataset from a tobacco company. The experimental results indicated that our methods have a significant advantage on total misclassification cost, high cost label recognition rate and average classification accuracy when compared with the cost-insensitive methods.
作者
张忠良
汤建国
乔丹娜
雒兴刚
赵亮
唐加福
ZHANG Zhongliang TANG Jianguo QIAO Danna LUO Xinggang ZHAO Liang TANG Jiafu(College of Information Science and Engineering, Northeastern University, Shenyang 110819, China Technology Center, China Tobacco Yunnan Industrial Co., Ltd., Kunming 650231, China)
出处
《中国烟草科学》
CSCD
北大核心
2016年第5期75-81,共7页
Chinese Tobacco Science
基金
国家自然科学基金面上项目"基于QFD和数据挖掘的卷烟产品叶组配方优化关键技术研究"(61273204)
关键词
分类算法
代价敏感
感官评估
神经网络
卷烟
classification algorithm
cost-sensitive
sensory evaluation
neural network
cigarette
