【作者】 李丽；

【导师】 张国印；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2010， 硕士

【摘要】 随着征信范围的扩大,用户缴费信用问题越来越受到政府和企业的重视。信用风险评估作为数据挖掘技术的一个新的应用领域尚未成熟,还需要不断深入地探索和研究。本文针对银行和证券方面信用风险研究比较成熟而用户缴费信用风险研究相对落后的情况,展开公共服务领域用户缴费信用风险评估研究。着力解决系统效率低、复用性差等问题。系统模型主要包括数据挖掘横向通用知识结构、数据挖掘算法建模推理支撑环境、数据挖掘调度平台、数据挖掘算法构件库四个部分。系统模型分离数据挖掘算法构件库,并基于挖掘算法的推理支撑环境研究,获得动态数据挖掘算法建模方案,通过设计挖掘算法路由配置机制,实现算法动态调度。数据挖掘算法动态配置的推理支撑环境设计,重点是知识库和推理机设计。一方面,根据算法方案配置需求设计了数据挖掘算法建模知识表达模型,基于建模知识基础完成算法路径的推理与解释;另一方面,推理机设计遵循数据描述约束及任务描述约束,结合相容和相斥运算得出数据挖掘算法的可行方案集合。公共服务领域用户缴费信用风险评价系统从运行和应用角度评价其质量。以电费缴纳业务的数据为例,挖掘用户非信用行为的主要特点,为企业建立预警机制提供参考。更多还原

【Abstract】 With the increasing acceptance of credit,the question of customer payment credit has already been paid attention to more and more by government and corporation. As a new application area, the study of credit ratings has not become mature yet. Many of problems need continued study.The study of credit risk evaluation methods in Banking and Securities industry has been mature, but the study of customer payment credit is relative backward industry. In such case, we conducted a study to customer payment credit risk evaluation methods.This paper design data mining model of customer payment credit risk evaluation, and address the problems of low efficiency and poor reusability. As a whole, the model of data mining system includes four parts: the lateral general knowledge structure of data mining; the inferential control support environment of data mining algorithm deployment construction; the platform of data mining management; the component libraries of data mining algorithm. The component libraries of data mining algorithm are designed independently. Depending on study of the inferential control support environment of data mining algorithm deployment construction, we can obtain the data mining algorithm deployment model. The paper proposes the mechanism achieve data mining algorithm router configuration for data mining algorithm dynamic scheduling.The inferential control support environment of data mining algorithm deployment construction is designed. The design of knowledge base and inference engine is a key problem. On the one hand, based on the requirement of studying data mining algorithm deployment , this paper designs data mining algorithm configuration construction knowledge presentation model, and completes the reasoning and interpretation of data mining algorithm paths. On the other hand, the design of inference engine follows the restriction of data description and task description. It achieves practicable data mining plan set by means of compatible arithmetic and repellent arithmetic.Customer payment credit risk evaluation system in public service field appraise the quality from point of view of operation and application. It find customer bad credit behavior pattern base on the customer information of electricity charges business, in order to provide consult and reference for building a credit warning mechanism in corporation.更多还原