【作者】 赫连佳鹏；

【导师】 景旭；

【作者基本信息】 哈尔滨理工大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 由于招投标研究具有十分可观的经济效益，所以在经济领域中招投标问题一直是人们关注的焦点。目前，投标报价决策研究方面的成果颇丰，总的来说是通过统计回归、模糊回归、专家知识系统以及基于事例分析等建模方法给出了可供投标人进行参考的报价方案。不难看出，支持这些成果的理论是一次价格密封竞标博弈理论。本文以一次价格密封竞标博弈理论为基础，用动态博弈观点看待投标人的报价形成过程，进行了投标人报价博弈模型的研究。概括如下：传统的报价决策模型对投标人之间的博弈进行分析时采用的假设是：投标人报价决策稳定不变，即与历史记录相同。显然，基于这样的假设分析仍然只是定性分析了投标问题中的博弈特征。为了实现对报价动态博弈过程的模拟，本文将理性预期思想与博弈理论相结合，指出投标人的报价是其综合实力的体现，即企业资源实力与“报价认知能力”共同决定投标人的报价。在资源实力固定的情况下，各个投标人在竞标博弈过程中的认知能力成为其能否中标的决定性因素。由于这个认知能力是随着环境在不断变化的，所以这里实现的是动态博弈的量化描述。考虑到投标人认知能力的典型特征，投标人的报价之间应该是严格的不可认知关系，所以，这里借用Luna提出的单层感知器不可“实现”异或函数的思想来映象投标人报价间的这种不可认知关系，从而建立投标报价博弈可计算模型。由于用带有两个逻辑函数的感知器来表示投标人的资源实力以及认知能力，有必要对Luna的基准表进行扩展。为了分析这种情况下可能出现的结果，需要找到Luna基准表背后的数学依据。这里利用几何分析和逻辑分析相结合的方法对这个基准表中两个逻辑函数组合的情况进行了深入分析，定性揭示了这个基准表背后的数学依据。值得指出的是，这样构造的投标人不是单层感知器结构，而是将逻辑函数作为中间层的双层感知器结构，但是由于从输入层到逻辑函数所在的中间层的权重不改变，所以需要给出这种情况下感知器参数调整算法，本文采用了Luna<WP=7>给出的算法来调整参数。算法以及交互规则的设计中都蕴涵了涌现宏观规律的隐秩序。进行仿真框架设计时引用了ERA方案，并在Swarm平台上对所建立的投标报价博弈可计算模型进行了仿真。设计方案结构的严格性确保了应用的清晰性，仿真结果给出了投标人的获胜概率以及这个概率在整体中所占的中标比例。最后，以两个投标人的竞标和三个投标人的竞标为例，我们对产生结果的隐秩序用本文给出的分析方法进行了详细的讨论，并结合一次价格密封竞标博弈理论对所有仿真结果所揭示的本质特征进行了解释，得到了合理的结论。本文力图从投标人的报价认知能力角度出发，对投标人报价形成的动态博弈过程进行量化分析和揭示，建立符合博弈思想、逼进客观实际的投标报价博弈模型，为一次价格密封竞标中“人的认知因素”的研究提供新的思路和方法，并为招标与投标仿真平台的建立创造条件。更多还原

【Abstract】 Actting as an important method of construction projiec, bid problem gives rise to attration. People gradually reconginze this researhch’ value.Since the bid method is so important that the theory about it is being developed quickly. Combines the thought of game-theory and the cognition of bidder’s quoted price, the first-price sealed game-theory based ,this paper investigates the cognition game model of bidder’s quoted price .The key problems as follows: First of all, while in the progress of the bidding making, its not reasonable to consider that the bidder’s quoted price would be one forever. The classed models analyzed the game phenomenon under the hypothesis that bidder’s quoted price should do so. Obviously ,this hypothesis is too far away form the reality. In order to conquer this weakness, we took on the research of the cognition of bidder’s quoted price while he in competitive environment. Second, considering the un-cognition between the bidders’ quoted price, we adopt the method of Luna which is using the single-lay perceptron to achieve XOR function. Just for this kind of net couldn’t do that ,so Luna map this un-cognition to all un-cognition phenomenon. Based above thoughts, we build the cognition game model of the bidder’s quoted price, and adopt the benchmark table of Luna to choose the ability of the bidders. By the method of geometry and logic analysis we bring out the theory under the benchmark table. On the other hand, to revise the parameters of our special double-lay perceptron, adopt the method put forward by Luna that is called gradient vector method. While we simulate this model, adopt the ERA scheme, and carry out on the Swarm software frame. In the result figure, we probe the win-PROB and the proportion figure of the bidder’s quoted price, and from those figures we choose the winner. Finally, illustrating the example of two bidders, and explaining the reason of hidder order’s production exactly, we apply the first-price sealed game to get the same result, open out the essential characteristic of the bid game theory. This paper make great efforts form the point of cognition of bidder’s quoted<WP=9>price, to profoundly research the progress of the quoted price how it comes into being. On the other hand, our work has no precedent ,so the success of this model would pioneer the research of the bid problem, bring new thought and methods for the bid researchers.In one word, our work has the profound value, and the model we’ve build move forward the bid research.更多还原