【作者】 唐亮贵；

【导师】 程代杰； 杨虎；

【作者基本信息】 重庆大学 ， 计算机软件与理论， 2009， 博士

【摘要】 信息技术、网络与服务型经济形态的迅速发展,使基于互联网环境的商务交易模式已逐渐进入实用化阶段,并成为企业核心竞争力乃至国家综合实力的重要体现。基于互联网的在线交易模型的研究与应用也因此得到了空前发展,并在传统商务交易模型向网络环境迁移方面取得了丰富的成果,它们主要集中在在线拍卖和招投标交易模型等领域。但是如何进一步将网络应用中信息服务的技术特征与电子商务特有的经济属性相结合,建立有效的电子商务微观市场结构、交易模型与运行机制,则是影响电子商务向纵深发展的关键问题。因此,开展电子商务市场结构及交易模型的研究是一项具有重要现实意义的工作。本文以实际工程应用项目为背景,以电子商务交易逻辑的模型化为核心展开研究,旨在进一步探索电子商务实用化、智能化和自动化的有效途径。具体包括以下工作:①分析了电子商务交易模型的经济技术特征,并有针对性地选择博弈论与信息经济学、MAS理论和智能计算方法等理论工具,建立了一种电子商务交易模型的计算架构,为构建高效实用的电子商务交易模型提供了理论与技术基础。②构建了一种有效的电子商务微观市场结构。论文针对市场主体、市场客体和运行机制三个基本要素,应用MAS理论,建立了多Agent协同商务市场模型(MACCM)和体系结构;针对多维商品属性空间难于实现交易过程中商品的快速检索与匹配,提出商品关键属性、撮合树和元模式组合特征等概念,并以此为基础构建了MACCM的微观逻辑结构和市场运行机制,从而可为电子商务交易模型的有效运行提供重要的基础设施和承载环境。③建立了多Agent撮合交易模型。为了实现市场交易的自动和有效匹配,论文运用市场博弈一般均衡理论,建立了多Agent撮合交易模型和算法;针对市场信息效率在电子商务交易中的特殊重要性,提出交易熵和交易熵链的概念,并分析了市场信息效率对价格波动的影响。该模型的建立,有利于构造合理的市场价格行为、优化市场微观结构和市场运行机制,能提高市场交易效率和市场效益,促使市场资源配置经济有效。④建立了多Agent动态竞标模型。多Agent动态竞标作为交易主体向市场提交买卖意向的必要环节,是构成有效市场环境的关键,因此论文根据交易Agent的个体行为特征,应用演化博弈理论,建立了支持向量机市场预测模型和神经网络随机反应均衡强化学习模型,设计了相应的算法和多Agent动态竞标流程。使交易Agent通过预测和学习,不断调整交易信念,规避市场风险,优化竞标策略,从而最大化其预期效用,可有效促进MACCM动态博弈的形成、市场价格的构造和各方收益的均衡。⑤建立了多Agent协商交易模型。协商是电子商务交易的重要阶段,针对MACCM中存在的潜在交易,论文应用Bayesian纳什均衡理论,提出了不妥协度的概念,构建了基于不妥协度的协商交易协议和Bayesian学习多Agent协商交易模型。使交易Agent通过有限轮协商,能够对协商对手的信念做出较为准确的判断,优化自身交易策略,从而进一步扩大市场成交量,提高市场效益。⑥最后对论文工作进行了总结,并对今后的研究方向进行了展望。论文还通过对部分轿车交易数据进行模拟实验,初步分析和验证了本文所建立的多Agent协同商务市场理论、模型和算法的技术可行性和经济有效性。更多还原

【Abstract】 With the rapid development of the information technology, the network and the form of service-economy, the commerce mode based on Internet environment gradually comes into practical phase, and becomes the enterprise’s core competence, and even to be the important aspect of the Comprehensive National Power. So the researches and applications about the online bargaining model based on Internet are unheard-of development, and the plentiful fruits are gained at the aspect of transplanting traditional trading models to the network circumstance, which almost aims at the fields of online auctions and biddings.But how further integrates the technical character of the information services in network applications with the unique economic attributes of the Electronic Commerce (EC), and to establish the effectual EC market microstructure, the trading model and the operational mechanism, which is the sticking point of the EC depth development. Therefore, to carry through the research about the electronic commerce market structure and trading model is practical and significant.Based on several actual application projects, this dissertation makes researches around with the EC’s trading logic modeling, for further exploring the effectual approaches to make EC become practical, intelligent and automation. The primary works and innovations are summarized as follows.First of all, this dissertation analyzes the technical and economic attributes of EC Trading Model (ECTM), and pertinently selects the theories of the game and information economics, the Multi-Agent Systems (MAS), and the intelligence computation, constructs an intelligence computational frame of the ECTM, so provides the important foundation of the theory and technology to establish the effective and practical ECTM.Second,according to three essential elements of market structure, this dissertation constructs an effectual EC market micromechanism.There establishes a Multi-Agents Cooperation Commerce Market model (MACCM) and its architecture by the theory of MAS. Aims at that it is difficult to implement the merchandise’s fast indexing and matchmaking in the process of trading with the high dimensional space of merchandise’s attributes, this dissertation puts forward the concepts of key attributes, matchmaking tree and the meta-pattern combination eigen, and constructs a micro logic structure and an operating mechanism of MACCM. So it can provide the important infrastructure and carrying environment for ECTM.Third,to implement the effective automatch of the transactions, and provide an effectual trading mechanism of EC market, this dissertation contrives a Multi-Agents Matchmaking Trading Model (MAMTM) by the theroy of general game equilibrium. Aims at the especial importance of the EC market information efficiency, the thesis puts forward the concepts of Trading Entropy (TE) and TE-Chains, analyzes its impact to quotation waves.These fruits are propitious to construct reasonable market price behavior, optimize the operational mechanism of the micro-market structure, boost up the trading efficiency and market benefits, and urge forward the economy effective collocation for market resources.Fourth, for referring the order to market, the multi-agents dynamic bidding is the vital element to form the effective market conditions, so this dissertation uses the evolvement game theory, and establishes a price forecasting model by the Support Vector Machine (SVM), contrives a Neuralnetwork Quantal Response Equilibrium Reinforcement Learning model (NQRERL), desgins the algorithm and the flow of the multi-agent dynamic bidding, which is according to the principle of minimizing experiential risk and the individual behaviors’character of trading agents.These works enable agents to adjust trading beliefs continuously, evade the market risks, and optimize bidding strategies by forecasting and learning, then maximize the anticipation avail and benefits. These are propitious to form the market game, pricing and equilibrium.Fifth, negotiation is the important phase of the EC trading, so this dissertation aims at the potential dealings, uses the theory of Bayesian Nash equilibrium, and puts forward the concept of UnCompromising Probability (UCP), constructs the multi-agent negotiation trading protocol, model and algorithm based on the UCP and the Bayesian learning equilibrium mechanism.By limited turn negotiations, the trading agent can infer the certain belief of the rival, consequently optimize their own strategy, so as to further increase the exchange quantity.Finally, the research work of this dissertation is summarized, and the future research direction is indicated. By some experiments and application examples about automobile transactions, there still analyzes and validates the technical feasibility and the economical validity of the MACCM theory, models and algorithms established in this dissertation.更多还原