报告题目：Data Mining for Social Security in E-Government Services

报 告 人：Chengqi Zhang, Professor（张成奇教授）

时    间：2011年9月21日下午3:40-5:10

地    点：中国人民大学信息楼4层学术报告厅

摘    要：

Social Security is a key government service for maintaining fairness and stability in any country. Due to the complexity of the services and the possibility of fraudulence, it is an ever-present problem which sometimes results in incorrect payments to beneficiaries. It is also difficult to collect debts from low-income beneficiaries which have resulted from incorrect payments. In practice, incorrect payments can be substantial, and this can impact on the perception and measurement of government performance. One possible solution for these problems is to use data mining technology to discover patterns which indicate fraud and incorrect payments, to optimise the assessment procedures and reduce the occurrence of incorrect payments. However, mining such data provides a challenge to existing Data Mining research in areas such as unbalanced data distribution and impact-targeted pattern mining.

In this talk, I will discuss how we can develop an innovative data mining methodology (activity mining) to reduce incorrect payments by optimising the assessment procedures. I will focus on how to discover impact-targeted activity patterns in huge volumes of unbalanced activity transactions. I will also discuss issues and prospects in mining high impact activities of exceptional behaviour from rare, dispersed and imbalanced data from governmental social security datasets.

In addition, I will also discuss how we can develop another new data mining methodology (combined mining) to enable faster and more efficient debt collection.  Here, I will focus on rule generation and interestingness measurements in combined association rule mining. In rule generation, the frequent itemsets are discovered among itemset groups to improve efficiency. New interestingness measurements are defined to discover more actionable knowledge.

The results we have achieved so far have been verified by the domain experts. The commercialised system based on the above research will be developed soon. Our experience can also be applied to other government services.

简   历：

张成奇教授于1991年获得澳大利亚昆士兰州大学（Queensland University）博士学位（Ph.D Degree），2002年澳大利亚墨尔本迪肯大学（Deakin University）理学博士后研究流动站出站（Doctor of Science from Deakin University which is the Higher Doctorate）。张成奇教授目前任澳大利亚悉尼科技大学量子计算与智能系统研究中心主任，澳大利亚国家人工智能理事会理事长，兼澳华科技协会会长。Zhang教授的研究兴趣包括多Agent系统，数据挖掘，以及相关的理论研究和应用开发工作。曾发表200多篇学术论文和6部学术专著，其中包括发表在Artificial Intelligence和IEEE/ACM 学报上的多篇高水平国际期刊论文。Zhang教授领导他的团队获得澳大利亚国家基金委300万美元的研究经费。曾应邀在12个国际学术会议上做过大会特邀报告。张教授是澳大利亚计算机学会院士 和 IEEE 计算机学会的高级会员。此外，还担任包括IEEE Transactions on Knowledge and Data Engineering在内的多个国际期刊的副编委，以及多个国际学术会议的大会主席，程序委员会主席，组织委员会主席等职务。包括International Conference on Knowledge Science, Engineering, and Management指导委员会主席，ICDM 2010大会主席，2017年国际人工智能联合大会（IJCAI）的Lacal Arrangement主席。张教授被选为理论科学、工程学和管理学方面的指导委员会主席。他自2004年以来，是PRICAI和PAKDD指导委员会的成员。张教授被邀请在国际会议上发表了六次著名的演讲。他也是IEEE计算机领域的资深成员。目前，张教授是国内许多知名院校的客座教授，如上海交大、中山大学等。更多的信息可访问他的主页http://www-staff.it.uts.edu.au/~chengqi/ 。