基于AdaBoost-SVM的葡萄酒品质分类模型优化设计
|
摘要
针对传统葡萄酒品质分类中低品质类葡萄酒样本识别率低的问题,提出一种基于集成支持向量机的葡萄酒品质分类优化算法.首先,通过"一对多"支持向量机实现多分类;其次,把支持向量机作为基分类器,反复训练支持向量机分类样本,通过AdaBoost得到多个支持向量机基分类器组合的强分类器,运用AdaBoost算法动态调整样本权值,适当提高低品质类样本权重,使低品质类中错判的样本代价增大,从而改进不平衡样本分类性能;最后,以Wine Quality数据集为研究对象,建立以多分类器优化集成为核心的葡萄酒品质分类模型.仿真结果表明,与传统的SVM算法相比,所提方法显著提高了低品质类葡萄酒分类精度.
Focused on the issue that traditional classification algorithms for wine quality classification have a low recognition rate to low-quality wines,an optimization algorithm based on ensemble Support Vector Machine(SVM)was proposed.Firstly,muti-class was accomplished by 1-against-the rest SVM;Secondly,SVM was repeatedly trained as weaker classifier and a strong classifier was gotten by grouping a number of base classifiers based on SVM.The sample weight were dynamically adjusted by using AdaBoost algorithm,the sample weight of low quality were appropriately increased,and then the cost of misjudge samples was also increased for improving classification performance of unbalanced datasets;Finally,the wine quality datasets of UCI database was taken as research object,the classification model of wines quality was established that using muti-classifiers optimal integration as the core.The simulation results show that compared with the standard SVM algorithm,classification accuracy of low quality wine was significantly improved based on AdaBoost-SVM.
Focused on the issue that traditional classification algorithms for wine quality classification have a low recognition rate to low-quality wines,an optimization algorithm based on ensemble Support Vector Machine(SVM)was proposed.Firstly,muti-class was accomplished by 1-against-the rest SVM;Secondly,SVM was repeatedly trained as weaker classifier and a strong classifier was gotten by grouping a number of base classifiers based on SVM.The sample weight were dynamically adjusted by using AdaBoost algorithm,the sample weight of low quality were appropriately increased,and then the cost of misjudge samples was also increased for improving classification performance of unbalanced datasets;Finally,the wine quality datasets of UCI database was taken as research object,the classification model of wines quality was established that using muti-classifiers optimal integration as the core.The simulation results show that compared with the standard SVM algorithm,classification accuracy of low quality wine was significantly improved based on AdaBoost-SVM.
引文
[1]何瑜.中国葡萄酒产业竞争力研究[D].杨凌:西北农林科技大学,2014.
[2]Baker A K,Ross C F.Sensory evaluation of impact of wine matrix on red wine finish:A preliminary study[J].Sensory Studies,2014,29(2):139-148.
[3]邵志芳.葡萄酒品质分析方法研究进展[J].中国酿造,2015,34(4):17-20.
[4]Paulo Cortez,Antonio Cerdeira,Fernando Almeida,et al.Modeling wine preferences by data mining from physicochemical properties[J].Decision Support Systems,2009,47(4):547-553.
[5]徐海涛.改进的近似支持向量机在葡萄酒质量鉴定中的应用[J].安徽农业科学,2010,38(29):16 105-16 106.
[6]刘延玲.新的Hopfield神经网络分类器在葡萄酒质量评价中的应用[J].价值工程,2012,35(2):181-182.
[7]Jose A.Seas,Bartosz Krawczyk,Michal Wozniak.Analyzing the oversampling of different classes and types of examples in multi-class imbalanced datasets[J].Pattern Recognition,2016,3(12):164-178.
[8]黄久玲.面向失衡数据集的集成学习分类方法及其应用研究[D].黑龙江:哈尔滨理工大学,2015.
[9]Zhong liang Zhang,Bartosz Krawczyk,Salvador Garcia,et al.Empowering one-vs-one decomposition with ensemble learning for multi-class imbalanced[J].Knowledge-Based Systems,2016,5(48):251-263.
[10]顾燕萍,赵文杰,吴占松.最小二乘支持向量机鲁棒回归算法研究[J].清华大学学报(自然科学版),2015,55(4):396-402.
[11]袁兴梅,杨明,杨杨.一种面向不平衡数据的结构化SVM集成分类器[J].模式识别与人工智能,2013,26(3):315-320.
[12]吕锋,李翔,杜文霞.基于MultiBoost的集成支持向量机分类方法及其应用[J].控制与决策,2015,30(1):81-85.
[13]李勇,刘战东,张海军.不平衡数据的集成分类算法综述[J].计算机应用研究,2014,31(5):1 287-1 291.
[14]李秋洁,茅耀斌.基于数据重平衡的AUC优化Boosting算法[J].自动化学报,2013,39(9):1 467-1 475.
[15]Ebenezer Owusu,Yong Zhao Zhan,Qi Rong Mao.An SVM-adaBoost-based face detection system[J].Journal of Experimental&Theoretical Artificial Intelligence,2014,26(4):477-491.
[16]李垒,任越美.基于改进AdaBoost集成学习的空间目标识别[J].计算机系统应用,2015,32(8):202-205.
[17]魏峻.一种有效的支持向量机参数优化算法[J].计算机技术与发展,2015,25(12):97-100,104.
[18]Paulo Cortez.Center for machine learning and intelligent systems[DB/OL].http://archive.ics.uci.edu/ml/datasets/Wine+Quality,2009-10-07.
[2]Baker A K,Ross C F.Sensory evaluation of impact of wine matrix on red wine finish:A preliminary study[J].Sensory Studies,2014,29(2):139-148.
[3]邵志芳.葡萄酒品质分析方法研究进展[J].中国酿造,2015,34(4):17-20.
[4]Paulo Cortez,Antonio Cerdeira,Fernando Almeida,et al.Modeling wine preferences by data mining from physicochemical properties[J].Decision Support Systems,2009,47(4):547-553.
[5]徐海涛.改进的近似支持向量机在葡萄酒质量鉴定中的应用[J].安徽农业科学,2010,38(29):16 105-16 106.
[6]刘延玲.新的Hopfield神经网络分类器在葡萄酒质量评价中的应用[J].价值工程,2012,35(2):181-182.
[7]Jose A.Seas,Bartosz Krawczyk,Michal Wozniak.Analyzing the oversampling of different classes and types of examples in multi-class imbalanced datasets[J].Pattern Recognition,2016,3(12):164-178.
[8]黄久玲.面向失衡数据集的集成学习分类方法及其应用研究[D].黑龙江:哈尔滨理工大学,2015.
[9]Zhong liang Zhang,Bartosz Krawczyk,Salvador Garcia,et al.Empowering one-vs-one decomposition with ensemble learning for multi-class imbalanced[J].Knowledge-Based Systems,2016,5(48):251-263.
[10]顾燕萍,赵文杰,吴占松.最小二乘支持向量机鲁棒回归算法研究[J].清华大学学报(自然科学版),2015,55(4):396-402.
[11]袁兴梅,杨明,杨杨.一种面向不平衡数据的结构化SVM集成分类器[J].模式识别与人工智能,2013,26(3):315-320.
[12]吕锋,李翔,杜文霞.基于MultiBoost的集成支持向量机分类方法及其应用[J].控制与决策,2015,30(1):81-85.
[13]李勇,刘战东,张海军.不平衡数据的集成分类算法综述[J].计算机应用研究,2014,31(5):1 287-1 291.
[14]李秋洁,茅耀斌.基于数据重平衡的AUC优化Boosting算法[J].自动化学报,2013,39(9):1 467-1 475.
[15]Ebenezer Owusu,Yong Zhao Zhan,Qi Rong Mao.An SVM-adaBoost-based face detection system[J].Journal of Experimental&Theoretical Artificial Intelligence,2014,26(4):477-491.
[16]李垒,任越美.基于改进AdaBoost集成学习的空间目标识别[J].计算机系统应用,2015,32(8):202-205.
[17]魏峻.一种有效的支持向量机参数优化算法[J].计算机技术与发展,2015,25(12):97-100,104.
[18]Paulo Cortez.Center for machine learning and intelligent systems[DB/OL].http://archive.ics.uci.edu/ml/datasets/Wine+Quality,2009-10-07.