摘要
为同时保证基分类器的准确性和差异性,提出一种基于聚类和AdaBoost的自适应集成算法.首先利用聚类算法将训练样本分成多个类簇;然后分别在每个类簇上进行AdaBoost训练并得到一组分类器;最后按加权投票策略进行分类器的集成.每个分类器的权重是自适应的,且为基于测试样本与每个类簇的相似性及分类器对此测试样本的分类置信度计算得到.实验结果表明,与AdaBoost,Bagging(bootstrap aggregating)和随机森林等代表性集成算法相比,该算法可取得更高的分类精度.
In order to ensure the accuracy and diversity of the base classifier at the same time,we proposed an adaptive ensemble algorithm based on clustering and AdaBoost.Firstly,the training samples were divided into multiple clusters by clustering algorithm.Secondly,AdaBoost training was performed on each cluster to get a set of classifiers.Finally,these classifiers were combined according to the weighted voting strategy.The weights of each classifier were adaptive,we calculated the similarity between the test samples and each cluster and got the test samples＇ classification confidence given by classifiers.The experimental results show that the algorithm can achieve a higher classification accuracy compared with representative ensemble algorithms such as AdaBoost,Bagging（bootstrap aggregating）and Random Forest.
作者
王玲娣
徐华
WANG Lingdi;XU Hua(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,Jiangsu Province,China)
出处
《吉林大学学报（理学版）》
CAS
CSCD
北大核心
2018年第4期917-924,共8页
Journal of Jilin University:Science Edition
基金
江苏省自然科学基金(批准号:BK20140165)
国家留学基金委项目(批准号:201308320030)
