基于A-D模型的K-means算法在通话异常客户挖掘中的应用

摘要

为能够利用海量的语音通信记录,高质量地对各种语音通信行为异常的客户(电信诈骗客户、广告客户等)进行聚类分析,设计构建了语音通信异常客户的行为特征模型,并基于该模型提出了一种语音通信行为异常客户的聚类分析算法.首先,通过分析客户的通话记录得出客户通话次数、接通率等通话行为特征,然后将AHP模型与DEMATEL方法融合,构建语音通信行为异常客户的行为特征模型(AHP-DEMATEL模型);其次,基于该模型提出了一种改进的K-means算法,实现根据语音通信记录对异常客户进行聚类分析.最后,使用真实数据进行了验证分析.结果表明,相较于其他类似算法,本文算法在多类型异常客户综合聚类分析和单类型异常客户聚类分析时,其性能都得到了较大幅度的提高.%In order to make use of massive voice communication records and cluster high-quality clients (telecom fraud clients,advertisers) with various kinds of voice communication abnormalities,a behavioral feature model of abnormal voice communication customers was designed and constructed.Based on the model,a clustering analysis algorithm for customers with abnormal voice communication behavior was proposed.First of all,by analyzing the call records of customers,the characteristics of call behaviors was got,such as the number of calls,call rates,and so on.Then AHP-DEMATEL model was constructed by blending the AHP model and DEMATEL method.Secondly,based on the model,an improved K-means algorithm was proposed to cluster the abnormal clients according to the voice communication records.Finally,the real data was used to verify the analysis.The results show that compared with other similar algorithms,the proposed algorithm improves the performance of multi-type abnormal customer comprehensive clustering analysis and single-type abnormal customer clustering analysis greatly.