航线网络需求非限化估计中的策略行为发现算法

A Strategic Behavior Discovery Algorithm for Demand Unconstraining Estimation in Airline Network

已有的优化类行为发现算法研究仅涉及短视型顾客,未考虑网络环境下的顾客策略行为。针对现有基于顾客偏好排序列表的网络型非参数离散选择模型,提出了考虑策略型顾客的偏好行为发现启发式算法。从定义顾客类型初始集和备选集开始,描述了受到顾客类型数量限制的模型参数极大似然联合估计问题,并通过广义拉格朗日函数将其转化为无约束问题,利用拉格朗日函数及其对偶函数梯度的性质确定顾客类型筛选条件。数值模拟表明,所提启发式算法能反映出顾客需求的网络替代效应,迭代并自动地生成符合筛选条件的短视型和策略型顾客偏好排序列表,其行为发现计算过程无须求解NP难问题,能更加有效地避免需求非限化估计过程对历史顾客初始需求的高估。

The existing researches on optimization-based behavior discovery algorithms only involve myopic customer types and do not consider strategic customer behavior in the network environment.In view of the nonparametric discrete choice model considering substitution effects in network based on rank-based preference lists,a preference behavior discovery heuristic algorithm for the strategic customer was proposed.Starting from the definition of the initial set and the alternative set of customer types,the problem of maximum likelihood joint estimation of model parameters limited by the number of customer types was described,which was transformed into an unconstrained problem by constructing a generalized Lagrange function.The customer type screening conditions were determined by using the gradient properties of the Lagrange function and its dual function.Numerical simulation shows that the proposed algorithm can reflect the substitution effects in the network of customer demand.The proposed algorithm is able to iteratively and automatically generate the preference ranking lists of both myopic and strategic customer types which meet the screening criteria.The market discovery process of the proposed algorithm does not need to solve the NP-hard problem.It is more effective to prevent overestimating the primary demand of historical customers in the procedure of demand unconstraining estimation.