【作者】 周丹；

【导师】 饶文碧；

【作者基本信息】 武汉理工大学 ， 计算机科学与技术， 2011， 硕士

【摘要】 在当今这个信息的世界,随着数据库技术和计算机网络技术的飞速发展,数据的存储从单个数据库发展到分布式数据库,数据呈现出一种指数级的爆炸增长方式。这种形式下迫切需要有效的方法来处理和挖掘日益庞大的分布式数据,而这一需求促进了分布式数据挖掘技术的发展。多Agent技术比较适合于对复杂的分布式问题进行求解。在基于多Agent的数据挖掘系统中,数据挖掘Agent分布在网络中不同地点的数据库上,按用户的要求进行数据挖掘,并将结果传送给用户,避免了数据库中数据的移动。因此,将多Agent技术应用于分布式数据挖掘系统中,具有很大的技术优势。论文主要研究了基于多Agent的分布式数据挖掘方法,建立了一个系统模型并设计了各模块的功能、结构和算法等,并针对电力负荷预测这一问题对原型系统进行了仿真实验。论文主要工作如下：(1)介绍了多Agent技术在分布式数据挖掘中的应用背景,从分布式计算角度说明了几种主要的分布式数据挖掘技术,着重描述了基于Agent和多Agent技术的分布式数据挖掘及其国内外研究现状。(2)研究了数据挖掘和Agent技术的相关理论知识,包括数据挖掘的定义、方法、基本原理以及Agent技术的概念、特征和结构等,分析了多Agent系统的组织结构和分布式数据挖掘的体系结构。(3)设计了基于AATP的多Agent分布式数据挖掘模型,构建了算法分析、任务预测和负载均衡子模型；从服务提供层、任务调度层和用户接口层说明了各模块的功能,并设计了主要子模块的结构和算法等,分析了主要的工作流程。此外,对该系统模型的关键技术以及主要特点等进行了说明和分析。(4)针对电力负荷预测这一问题,对基于AATP的多Agent分布式数据挖掘模型进行了仿真实验和代码实现。论文采用欧洲智能技术网络组织提供的欧洲某些地区的电力负荷等相关真实数据,利用基于AATP的多Agent分布式数据挖掘模型系统进行处理,并对预测的负荷数据和预测性能做出了分析和对比,说明了该原型系统在电力负荷预测问题上的有效性和可行性。更多还原

【Abstract】 Today, the world is surrounded by a large amount of information. With the rapid development of database and computer network technology, large amounts of data are stored in many distributed databases instead of a single database, and data is growing in an index series explosion mode. This situation urgently needs an effective way to deal with and mining these increasingly large distributed data. And the demand promotes the development of Distributed Data Mining technology.The Multi-Agent technology is suitable for the complex distributed problems. In Data Mining System based on Multi-Agent, Agents on different database will be working according to the user’s requirements, and transferring the result to the users. In this way, the large amounts of data in every database avoid being moved. So there are advantages of using the Multi-Agent technology in the Distributed Data Mining Systems. This thesis mainly studies Distributed Data Mining methods based on Multi-Agent, establishes a system model and design the function, structure and algorithm of each module. A simulated experiment is done that building a prototype system which in order to handling the problem of power load forecasting. The main works of the thesis are as follows:(1) The thesis introduces the research background of the Multi-Agent technology that been used in Distributed Data Mining, and narrates several major Distributed Data Mining technology from the aspect of distributed computing technology. Especially, the Distributed Data Mining based on Multi-Agent technology and the situations of the domestic and international research are illustrated in this thesis.(2) The thesis researches the basic theories on data mining and Agent technology, including the definition, methods, basic principle of data mining and the concept, characteristics and structure of Agent. The organizational structure of Multi-Agent system and the architecture of Distributed Data Mining are discussed.(3) A Distributed Data Mining model based on Multi-Agent is given. Three main Agents:the Algorithm Analysis Agent, the Task Prediction Agent, and the Load-Balancing Agent are designed in the model. The system model is divided into three layers. They are the Service layer, the Task-scheduling layer and the User Interface layer. Every layer has themselves functions. The thesis also designs the structure and algorithms for each main module and analyzes the main work flow. In addition, the key technologies of system model and the main characteristic are explained and analyzed.(4) A simulation experiment about the Multi-Agent DDM model based on AATP is done, which is used for solving the problem of the power load forecasting. The experiment processes the real power load data and other relevant data of certain areas that provided by the European Intelligence Technology Network organization. At last, the thesis analyzes and contrasts the forecast load data and forecast performance, which shows the effectiveness and feasibility of this prototype system in power load forecasting problems.更多还原