二型模糊决策方法及其在个性化推荐中的应用

发布时间：2018-01-04 14:35

  本文关键词：二型模糊决策方法及其在个性化推荐中的应用　出处：《东南大学》2016年博士论文　论文类型：学位论文

  更多相关文章： 二型模糊集 信息集成 多属性决策 粒计算 个性化推荐

【摘要】：目前,基于大数据驱动的管理与决策方法研究,是管理科学领域研究的热点问题。由于决策环境和决策行为的复杂性,在大数据背景下如何根据用户的行为偏好和知识发现,实现对用户的个性化推荐,以及如何将多属性决策方法与个性化推荐系统进行有机结合,研究基于多属性决策理论的个性化推荐系统就具有重要的理论价值与现实意义。本论文将主要利用二型模糊集对语言和语义信息的强大处理能力,以二型模糊信息集成与决策方法为切入点,结合粒计算的相关方法和技术,研究基于二型模糊决策方法的个性化推荐模型,为大数据背景下的复杂、动态、信息不完全个性化推荐问题研究提供新的思路和解决方案。论文主要研究内容如下：(1)基于信息集成理论,提出了基于Maclaurin对称平均的区间二型模糊信息集成算子,并给出了其对偶形式及指数扩展形式。研究了其参数单调性,并指出相较于现有的区间二型模糊信息集成算子,区间二型模糊Maclaurin对称平均算子能够柔性地处理具有多重关联关系的区间二型模糊信息的集成问题。在此基础上,提出了一种处理区间二型模糊决策问题的方法,并将其应用于中国科技论文在线的论文评审推荐系统中。为研究具有关联关系的二型模糊决策问题提供新的方法。(2)基于多属性决策理论,分别从排序方法、效用模型和优化模型三个方面研究了基于区间二型模糊信息的多属性决策方法。首先,针对二型模糊集中的—个难点问题：排序问题。给出了一种基于三种初等平均的组合排序值方法。并且从数学的角度证明了该排序方法不但满足线性序(全序)关系,而且还满足admissiable序关系。在此基础上,进一步研究了基于组合排序值的区间二型模糊决策方法。然后,基于行为决策理论,借鉴行为经济学中的柔性三参数(FTP)效用函数,给出了基于FTP效用函数的二型模糊OWA算子,并针对大规模复杂决策问题,提出了基于模糊聚类的区间二型模糊多属性决策方法。进一步将前景理论与经典的VIKOR方法进行结合,研究了基于动态参考点的区间二型模糊行为决策方法,并将其应用于高新技术风险投资评估推荐系统中。最后,将多目标优化中的LINMAP(多维线性规划)方法扩展到了区间二型模糊环境下,研究了决策者偏好的提取方法,建立了一系列属性权重信息不完全时的优化决策模型,并将其运用在了手机购买的个性化推荐中。这些新方法的提出,进一步丰富了二型模糊决策方法的理论体系,同时也扩展了二型模糊决策理论在处理个性化商务推荐的应用范围。(3)基于粒计算理论,研究了个性化推荐中的评分矩阵的稀疏性问题。以粒计算方法为切入点,建立了以Coverage和Specificity准则为核心的协同优化模型。提出了求解该模型的智能优化算法。在一定程度上克服了现有的基于矩阵分解和变分优化方法所带来的高计算复杂性,为解决个性化推荐中的瓶颈问题提供了新的研究手段和方法。(4)基于二型模糊决策方法研究了个性化推荐模型。以两种新的多属性决策方法BTW和MULTIMOORA为基础,结合最优信息粒建模的思想,提出了基于多属性协同过滤和内容的混合推荐模型。对于推荐模型求解中的参数设置问题,研究了个性化、差异化的参数设置方法。本文中所提到的方法都在二型模糊多属性决策问题中得到了应用。相关研究成果在理论层面上可以进一步丰富和完善基于二型模糊信息的决策理论与方法；在应用层面上可以为电子商务的个性化推荐提供新的工具与方法。
[Abstract]:At present, the research of management and decision making method based on data driven, is a hot issue in the field of management science research. Because of the complexity of the decision environment and decision-making behavior, in the context of large data according to user preferences and knowledge discovery, realize personalized recommendation to users, and how to use the multi attribute decision making method and personalized recommendation system by combining the research of personalized recommendation system based on multi-attribute decision theory has important theoretical value and practical significance. This paper will mainly use the type two fuzzy sets the powerful processing ability of language and semantic information, to type two fuzzy information integration and decision method as the starting point, combined with the relevant methods of granular computing and technology study of type two fuzzy decision making method based on Personalized Recommendation Model for large data under the background of complex, dynamic, personalized recommendation problem with incomplete information The research provides new ideas and solutions. The main contents of this thesis are as follows: (1) based on the theory of information integration, puts forward the interval Maclaurin symmetric mean type two fuzzy operator based on information integration, and gives its dual form and index extended form were studied. The parameters of the single tone, and points out that compared with the existing range of two fuzzy information aggregation operator, interval type two fuzzy Maclaurin symmetric mean operator can flexibly deal with the multiple correlation interval type two fuzzy information integration problems. On this basis, put forward a method of processing interval type two fuzzy decision problems, and its application in China sciencepaper online paper review recommendation system. Provide a new method of type two fuzzy decision problem as the research has a relationship. (2) based on multi-attribute decision-making theory, respectively from the sorting method, the utility model and optimization Three aspects of research model of multi attribute decision making method based on interval type two fuzzy information. Firstly, according to the type two fuzzy sets is a difficult problem: scheduling problem. A value of three methods of combination of elementary average sorting based on given. And from the angle of mathematics proves that the method can not only satisfy the linear order sequence (total order), but also meet the admissiable relation. On this basis, further research on the fuzzy decision making method based on interval value ranking combination type two. Then, the behavioral decision theory based on reference number of flexible three parameters in Behavioral Economics (FTP) utility function, gives the fuzzy OWA operator based on FTP utility function type two, and for large and complex decision problem, we propose a fuzzy multi attribute decision making method based on fuzzy clustering interval two. Further VIKOR method combining with classic prospect theory, was researched. From the dynamic point of reference interval type two fuzzy behavior decision method, and its application in the high-tech investment risk assessment recommendation system. Finally, the LINMAP in multi-objective optimization (multidimensional linear programming) method is extended to the interval type two fuzzy environment, extraction method studied the preference of decision makers, established a a series of incomplete information on attribute weights and the optimal decision-making model and its application in personalized recommendation to buy mobile phone. The new method, further enrich the theoretical system of type two fuzzy decision making method, but also extends the type two fuzzy decision theory in the application scope of processing personalized business recommended (3). Based on Granular Computing Theory and sparsity problem of personalized recommendation in the score matrix. Based on granular computing method as the starting point, based on the Coverage and Specificity standards as the core of the collaborative optimization model. The intelligent optimization algorithm to solve the model. In a certain extent overcome the high computational complexity of matrix decomposition and variational optimization method based on the existing brought, provide a new research means and methods for solving the bottleneck problem in personalized recommendation. (4) type two fuzzy decision making method based on personalized recommendation model. Two kinds of multi attribute decision making method of new BTW and MULTIMOORA as the foundation, combined with the optimal particle information modeling method, put forward the recommendation model for hybrid multiple attribute collaborative filtering and content-based recommendation model. The parameters for the solution of the problem of setting, personalized, parameter setting method has been applied to differentiation. The method mentioned in this paper are type two fuzzy multiple attribute decision making problems. The relevant research results in theory can further enrich and improve the type two fuzzy information based on decision theory. On the application level, it can provide new tools and methods for the personalized recommendation of e-commerce.

【学位授予单位】：东南大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.3;F274
，

本文编号：1378830