基于社交网络的信用评估模型的研究与实现

发布时间：2018-01-12 19:03

本文关键词：基于社交网络的信用评估模型的研究与实现　出处：《中北大学》2017年硕士论文　论文类型：学位论文

【摘要】：互联网科技的蓬勃发展使得社交网络已成为人们之间相互沟通、社交等活动的重要手段。虚拟世界与现实世界已在不知不觉中相互交融并相互影响,过往人和人之间面对面的信任关系已延伸至网络。社交网络在给人们的生活带来便利和趣味的同时,也引发了很多的困扰和威胁。社交失信事件时有发生,社交用户的信用情况愈来愈引起人们的关注。如何监管网络环境,规范、约束网民的行为成为目前需要面对的重大问题,这个问题的解决对我国网络诚信监管建设具有重要的意义。论文通过对现有信贷个人信用评估模型研究发现,单一的信用评估模型已趋向成熟,很难再有突破和扩展;而且很多的研究证明单一的信用评估模型弊端凸显,通过集成学习,可以使单一模型进行互补,极大的改善单一模型的预测精度和稳定性等性能,但是当单一模型数量过大时,会出现样本学习时间过长,导致信用评估效率下降,也加大了机器存储空间的要求。针对这些问题,“选择性集成”的思想又被提出。选择性集成方法目前已经被应用到很多的领域,并且均取得了较好的研究成果,但是在个人信用评估领域,研究成果相对来说还是较少,因此将选择性集成方法应用于个人信用评估领域有很大的研究空间。针对社交网络存在的网络诚信问题,论文以新浪微博为研究对象,把“选择性集成”的思想引入到微博信用评估领域,论文的主要工作内容如下:(1)结合微博平台特点及现有评估指标体系存在的问题,重建了信用评估指标体系,通过对比实验验证了该指标体系的有效性;(2)将选择性集成学习的思想引入到微博用户信用评估领域,提出一种基于K-means聚类和萤火虫群优化选择的KGSO选择性集成算法;(3)将KGSO选择性集成算法应用到微博用户信用评估中,并通过对比实验验证了KGSO算法的有效性和优越性;
[Abstract]:Rapid development of Internet technology makes the social network has become an important means of communication between people, social and other activities. The virtual world and the real world has imperceptibly interaction and mutual influence between the past, people face-to-face trust has been extended to the network. The social network bring convenience and fun at the same time to the people life, also caused a lot of problems and threats. Social credit events have occurred, social networking users credit has attracted more and more attention. How to regulate the network environment, norms, constraints the behavior of Internet users become the major problem facing the solution of this problem has important meaning to the network supervision of construction in China. Based on the existing credit evaluation model of credit evaluation model, the single is mature, very difficult to have a breakthrough and expansion; But many studies prove the obvious shortcomings of credit evaluation model through a single, integrated learning, can make a single model are complementary, improve the performance of the single model prediction accuracy and stability greatly, but when the single model number is too large, there will be a sample learning time is too long, resulting in a decline in credit evaluation efficiency, but also increased the machine the storage space requirements. To solve these problems, "selective integration" ideas have been proposed. Selective ensemble method has been applied to many fields, and have achieved good results, but in the field of personal credit evaluation, research is still relatively small, so the application of selective integration method in personal credit evaluation the field has a lot of research space. Aiming at the problem of network integrity of social networks exist, the Sina micro-blog as the research object, the "selective set "The idea is introduced to micro-blog credit evaluation, the main contents of this paper are as follows: (1) according to the existing micro-blog platform features and the existing evaluation index system, the reconstruction of the credit evaluation index system, through the experimental results verify the validity of the index system; (2) selective ensemble learning is introduced into micro-blog user credit evaluation, this paper puts forward a K-means clustering and glowworm swarm optimization algorithm based on selective ensemble selection of KGSO; (3) KGSO selective ensemble algorithm is applied to the micro-blog user credit evaluation, and through the experimental results verify the validity and superiority of KGSO algorithm;

【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【相似文献】