【作者】 路阳；

【导师】 许少华；

【作者基本信息】 大庆石油学院 ， 计算机应用技术， 2005， 硕士

【摘要】 目前,软件Agent技术已被广泛应用于各个领域,尤其是在WWW网上信息检索、电子商务、网络管理等的应用。这对 Agent的适应性提出了更高的要求。适应性也是衡量这个软件系统是否是 Agent系统的主要特性之一。如何让 Agent 能够更有效、更准确地找到所需信息,并对这些信息进行过滤、分类,根据Agent自身具有的知识进行推理,然后根据决策计划行动并执行,是计算能否成功的关键。所有这些都要求 Agent具有很强的自适应能力。　因此我们基于 Agent的 BDI模型设计和实现了一个信息 Agent系统,它能根据用户定义的要计算目标,自主去适应Internet环境,不断的学习,然后从WWW服务器文件中发现并分类提取与目标相关的信息。该模型集成了 SOM 神经网络,遗传算法,信息过滤等算法,使 Agent具有更好的学习能力,从而可以使 Agent更加智能化。　信息 Agent体系结构包含以下几个基本模块:　 用户界面用以接受用户信息输入或输出信息给用户。　 通信接口用来与其它软件Agent或应用进行通信。　 感知模块对输入信息进行过滤与分类。　 推理模块根据Agent自身知识对信息进行推理。　 决策模块对推理结果进行评价和决策。　 计划模块根据决策制订行动计划。　 执行模块按照计划执行动作。　 知识库对推理、决策、计划等提供支持。　信息 Agent系统中用户界面 Agent 采用 SOM神经网络来学习用户的偏好,兴趣,网络经过训练实现文档向量到系统输出的映射,把代表不同文档的特征向量映射为用户的不同目标,从而实现区别用户不同目标的任务。搜索 Agent中引入遗传算法,帮助用户在Internet环境中寻找潜在的信息。　本文结合“油气管道信息服务平台”项目,提出了利用“信息 Agent 系统”来提高管道文献检索的智能性。在项目的实践过程中,通过文献数据库及偏好知识库的构建及网络蜘蛛、索引器等应用程序的实现,增强了信息 Agent系统主动适应Internet动态环境与自学习、自适应能力。本论文提出的方法取得了较好的效果,虽然还存在着缺陷,但基本解决了文献检索中信息个性化服务的问题,提高了管道文献信息查全率与查准率。更多还原

【Abstract】 At present,Software Agent technology has already been widely used in each field,especially the application in WWW online information retrieval,e-commerce,network management,etc..This has put forward higher request to adaptability of Agent.It is one of the main characteristics of Agent system that adaptability weighs some software system.How to let Agent find necessary information more effective,more accurately,filter to and classfy them,carry on reasoning according to the knowledge that Agent oneself has,then plan to take action and carry out according to the policy,which is the key to calculating that could be successful.All these require Agent has strong adaptive capacity. So we designed and realized a information Agent system based on BDI model,it can calculate the goal according to what users defined,go to adapt to Internet environment independently,constant study,then find and classify the information drawn from WWW server file.The system integrated SOM neural network ,hereditary algorithm and algorithms of filtering niformation.Thus the agent has more intelligent and better learning abilityis. Information Agent system structure includes the following modules: User’s interface is used to accept user’s information and output information to users. Communication interface is used for communicating with other software Agents. The perceiving module carries on filtering and classify the information. The reasoning module carries on reasoning to information according to agent’s knowledge. The decision module appraises the reasoning result and makes policy. The planning module makes the action plan according to above policy. The carrying module takes action according to schedule. The knowledge base supports to reasoning,decision,plan,etc.. User interface Agent adopts SOM neural network to study user’s partiality in the information Agent system,and the network realizes mapping from document vector to system’s output,then the network will know user’s different goals from different vectors,so distinguished the user’s different goals.We introduced the hereditary algorithm in the searching Agent and helped users look for potential information. This paper based on “pipeline information service platform of oil gas”,proposed making use of “Information Agent System”to improve searching intelligent.In the course of developing the project,we realized document database,knowledge base and network spider,etc.applications,then strengthen dynamic adaptive capacity with studying by oneself.All these methods have made better results and the defect exists,but solved personalized problem and raised the recall ratio and precision ratio.更多还原