Behavior analysis on shanghai private car license auction market with statistical learning methods

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.