基于OVO分解策略的智能卷烟感官评估方法
|
摘要
针对智能卷烟感官评估系统中涉及的多分类问题,采用"一对一"(one-versus-one,OVO)分解策略将复杂的多分类问题分解成多个易于处理的二分类子问题,然后针对这些子问题分别建立二值分类器,最后采用一定的聚合策略将二值分类器组合成多类分类器.此外,分别采用基于动态分类器选择和基于距离相对竞争力加权法对OVO中的冗余二值分类器进行处理,从而降低其对OVO系统的消极影响.为了验证所采用的方法在智能卷烟感官评估中的有效性,采用国内某烟草公司提供的数据集进行对比实验.实验结果表明,在智能卷烟感官评估中基于OVO分解策略的多分类方法比传统方法具有更优的分类性能.
Intelligent cigarette sensory evaluation system involves multi-class classification problems. The one-versus-one(OVO) decomposition strategy was employed to divide the multiclass classification problem into several easier-to-solve binary sub-problems. Then binary classifiers were established for these sub-problems. Finally,an aggregation strategy was adopted to combine the binary classifiers to be a multi-class classifier. In addition,dynamic classifier selection for OVO strategy(DCS-OVO) and distance-based relative competence weighting for OVO strategy(DRCW-OVO) were used to reduce the negative effect of the non-competent classifiers. In order to verify the effectiveness of the employed method in intelligent cigarette sensory evaluation,the experimental comparison by using the dataset from a Chinese tobacco company was carried out. The results indicate that the OVO decomposition strategy outperforms the classical methodology in intelligent cigarette sensory evaluation.
Intelligent cigarette sensory evaluation system involves multi-class classification problems. The one-versus-one(OVO) decomposition strategy was employed to divide the multiclass classification problem into several easier-to-solve binary sub-problems. Then binary classifiers were established for these sub-problems. Finally,an aggregation strategy was adopted to combine the binary classifiers to be a multi-class classifier. In addition,dynamic classifier selection for OVO strategy(DCS-OVO) and distance-based relative competence weighting for OVO strategy(DRCW-OVO) were used to reduce the negative effect of the non-competent classifiers. In order to verify the effectiveness of the employed method in intelligent cigarette sensory evaluation,the experimental comparison by using the dataset from a Chinese tobacco company was carried out. The results indicate that the OVO decomposition strategy outperforms the classical methodology in intelligent cigarette sensory evaluation.
引文
[1]王强,李孟军,陈英武.卷烟配方数据挖掘技术研究进展[J].中国烟草科学,2007,28(4):14-17.(Wang Qiang,Li Meng-jun,Chen Ying-wu.Research progress in data mining technology on cigarette formulation[J].Chinese Tobacco Science,2007,28(4):14-17.)
[2]邵惠芳,许自成,李东亮,等.基于BP神经网络建立烤烟感官质量的预测模型[J].中国烟草学报,2011,17(1):19-25.(Shao Hui-fang,Xu Zi-cheng,Li Dong-liang,et al.The establishment of BP neural network based models for predicting tobacco leaf sensory quality[J].Acta Tabacaria Sinica,2011,17(1):19-25.)
[3]赵青松,李兴兵,唐小松.基于支持向量机的烟叶感官品质评价[J].计算机工程与应用,2007,43(10):236-240.(Zhao Qing-song,Li Xing-bing,Tang Xiao-song.Tabacum sensory evaluation based on the support vector machine[J].Computer Engineering and Applications,2007,43(10):236-240.)
[4]Galar M,Fernández A,Barrenechea E,et al.An overview of ensemble methods for binary classifiers in multi-class problems:experimental study on one-vs-one and one-vs-all schemes[J].Pattern Recognition,2011,44(8):1761-1776.
[5]Hüllermeier E,Vanderlooy S.Combining predictions in pairwise classification:an optimal adaptive voting strategy and its relation to weighted voting[J].Pattern Recognition,2010,43(1):128-142.
[6]Huhn J C,Hüllermeier E.FR3:a fuzzy rule learner for inducing reliable classifiers[J].IEEE Transactions on Fuzzy Systems,2009,17(1):138-149.
[7]Hüllermeier E,Brinker K.Learning valued preference structures for solving classification problems[J].Fuzzy Sets and Systems,2008,159(18):2337-2352.
[8]Orlovsky S.Decision-making with a fuzzy preference relation[J].Fuzzy Sets and Systems,1978,1(1):155-167.
[9]Fernández A,Calderón M,Barrenechea E,et al.Solving multiclass problems with linguistic fuzzy rule based classification systems based on pairwise learning and preference relations[J].Fuzzy Sets and Systems,2010,161(23):3064-3080.
[10]Galar M,Fernández A,Barrenechea E,et al.Dynamic classifier selection for one-vs-one strategy:avoiding non-competent classifiers[J].Pattern Recognition,2013,46(12):3412-3424.
[11]Galar M,Fernández A,Barrenechea E,et al.DRCW-OVO:distance-based relative competence weighting combination for one-vs-one strategy in multi-class problems[J].Pattern Recognition,2015,48(1):28-42.
[2]邵惠芳,许自成,李东亮,等.基于BP神经网络建立烤烟感官质量的预测模型[J].中国烟草学报,2011,17(1):19-25.(Shao Hui-fang,Xu Zi-cheng,Li Dong-liang,et al.The establishment of BP neural network based models for predicting tobacco leaf sensory quality[J].Acta Tabacaria Sinica,2011,17(1):19-25.)
[3]赵青松,李兴兵,唐小松.基于支持向量机的烟叶感官品质评价[J].计算机工程与应用,2007,43(10):236-240.(Zhao Qing-song,Li Xing-bing,Tang Xiao-song.Tabacum sensory evaluation based on the support vector machine[J].Computer Engineering and Applications,2007,43(10):236-240.)
[4]Galar M,Fernández A,Barrenechea E,et al.An overview of ensemble methods for binary classifiers in multi-class problems:experimental study on one-vs-one and one-vs-all schemes[J].Pattern Recognition,2011,44(8):1761-1776.
[5]Hüllermeier E,Vanderlooy S.Combining predictions in pairwise classification:an optimal adaptive voting strategy and its relation to weighted voting[J].Pattern Recognition,2010,43(1):128-142.
[6]Huhn J C,Hüllermeier E.FR3:a fuzzy rule learner for inducing reliable classifiers[J].IEEE Transactions on Fuzzy Systems,2009,17(1):138-149.
[7]Hüllermeier E,Brinker K.Learning valued preference structures for solving classification problems[J].Fuzzy Sets and Systems,2008,159(18):2337-2352.
[8]Orlovsky S.Decision-making with a fuzzy preference relation[J].Fuzzy Sets and Systems,1978,1(1):155-167.
[9]Fernández A,Calderón M,Barrenechea E,et al.Solving multiclass problems with linguistic fuzzy rule based classification systems based on pairwise learning and preference relations[J].Fuzzy Sets and Systems,2010,161(23):3064-3080.
[10]Galar M,Fernández A,Barrenechea E,et al.Dynamic classifier selection for one-vs-one strategy:avoiding non-competent classifiers[J].Pattern Recognition,2013,46(12):3412-3424.
[11]Galar M,Fernández A,Barrenechea E,et al.DRCW-OVO:distance-based relative competence weighting combination for one-vs-one strategy in multi-class problems[J].Pattern Recognition,2015,48(1):28-42.