上海私车额度拍卖系统行为统计学习分析

Behavior Analysis on Shanghai Private Car License Auction Market with Statistical Learning Methods

私车额度拍卖是政府为抑制交通需求所采取的一种经济控制手段.本文选用上海私车额度拍卖市场中车牌投放量、竞买人数和牌照均价三个关键变量作为研究对象,根据拍卖市场所表现出的整体运行规律,采用针对性的统计学习方法,深入分析它们之间存在的复杂的耦合互动关系.通过主成分分析发现,车牌投放量、竞买人数和牌照均价之间存在近似的线性约束关系,车牌投放量在其中起到了关键的制约作用,显著影响了竞买人数和牌照均价两个变量.同时,本文应用概率型自组织神经网络算法,根据有限的试验样本,精确地估计出这三个变量的各种概率分布函数,建立了准确的系统概率分布模型,从而客观、量化地描述出拍卖市场的系统行为.本文的研究比较精确地刻画了各市场要素之间的定量关系,可为确定车牌投放量、优化拍卖规则,以及规范市场秩序提供决策参考.

The quota auction of private car licenses is one of the market-based regulatory measures used by the local government to control the vehicle demand. According to the operating rules of the auction market,this paper selects three key elements： the released quotas, the number of bidders and the average prices as analytical variables. It studies the system behaviors and the coupling interactions among the key elements with statistical learning methods. By using the principal component analysis, the paper proposes a linear constraining relationship among the key elements, which quantitatively shows that the released quotas play an important role on effecting and restricting the other two elements. Meanwhile, the probabilistic self-organizing network is used to accurately estimate the joint-probability and the conditional-probability distributions of these elements from the limited training samples. Finally, a probability distribution model is developed to describe the behavior of the auction market objectively and quantitatively. Based on the precise characterization of the relationship between the key elements, the paper provides some constructive advice on determining the released quotas, optimizing the auction rules and regulating market order.