Agent网络法——一种多元商务智能技术融合方法

为了融合粗糙集、模糊逻辑、决策树、群决策、蚁群算法、数据挖掘、遗传算法、人工神经网络等异质商务智能（BI）技术,基于复杂网络和多Agent,提出了一种称为Agent网络法的融合方法,该方法包括3大机制：自治主体,网络集聚的智能资源集成机制;“竞争上岗,择优合作”的智能资源利用机制;“绩效评估,优胜劣汰”的智能资源优化机制.基于该方法构建的BI系统是一个以众多Agent为节点的复杂网络,其中的每个Agent都是一个独立的智能单元,代表BI技术中的一个计算模型或算法.3大机制实现了多元BI技术两个层次的融合：新的Agent不断加入,被实践证明低价值的旧Agent不断剔出,网络中始终聚合着所有高价值的Agent,从而在系统结构层面上实现优秀BI技术的全面融合；借助谈判、招标、拍卖等竞争性合作机制,每次响应任务时,选出的是网络中最能胜任当前任务的Agent组合,从而在系统应用层面上实现 BI技术的动态融合.

Agent-network:a fusion method for heterogeneous business intelligence technologies

In order to fuse heterogeneous business intelligence (BI) technologies such as rough set, fuzzy logics, decision tree, group decision, swarm algorithm, data mining, genetic algorithm, artificial neural networks, a complex network and multi-Agent based  method called as Agent-network was put forward. The method includes three basic mechanisms: intelligence resource (IR) aggregating mechanism based on autonomous Agent and complex network, IR using mechanism based on “competing for post” and “selecting the best for cooperation”, IR optimizing mechanism based on performance evaluation and “survival of the fittest”. A BI software system based on the method is a complex network composed of many Agents as nodes. Each Agent is a unit of BI resource which represents a computing model or an algorithm. By means of the three mechanisms of Agent-network, a BI system can realize two levels of BI technology fusion. One is called as fusion on system structure: new Agents are added into the network continuously and less value existing Agents are deleted from it, all valued Agents (on behalf of all excellent BI technologies) are aggregated in the system. Another is called as fusion on system application: by using some marketing mechanism such as negotiation, bidding and auction, the system always can find an optimal Agent portfolio to respond to a specific service request of user．