以模擬方法分析網路營收管理模式之重複求解議題

Abstract

本研究針對航空網路型態營收管理(RM)模式之重複求解議題進行探討。經典的動態規劃(DP)模式可被用以產生網路型態營收管理問題之最佳動態機位存貨控管決策，然而當路網規模擴大時，動態規劃模式由於眾所皆知的維度詛咒問題而無法進行求解。因此，有許多研究以靜態公式為基礎發展啟發式解法以產生近似的控管決策，此類啟發式解法多需隨著可接受訂位時間段的演進，而必須進行重複求解以處理機位控管決策問題之動態特性。本研究藉由模擬實驗的方式，以一最佳控管決策可進行求解之小規模路網問題為基礎，探討目前較普遍被使用的靜態線性規劃模式競價法其重複求解頻率的影響，此外，亦考慮一以參數化函數進行機位控管之決策方法，因參數化函數法在同樣的問題下，僅需求解一次即可進行機位控管，故模擬參數化函數法以作為重複求解頻率之評估基準。根據模擬實驗之結果，參數化函數法之表現非常接近最佳機位控管決策模擬結果，至於競價法機位控管決策，若欲達到與最佳控管決策模相當之結果，則需要較高之求解頻率，且模擬結果亦顯示，當競價法更新頻率不足時，其模擬收益結果明顯下降。

This research focuses on the re-solving issue of the revenue management (RM) problem for the airline network. The classical dynamic programming (DP) model can be used to derive the optimal dynamic control for the seat inventory in the network RM problem. However, when the network scale increases, the DP model can not be solved due to the well-known curse of dimensionality. Therefore, most heuristic methods were developed to generate the approximate control decisions based on the static formulations, which in general need to be re-solved within the booking horizon so as to handle the dynamic features of the problems. By a simulation experiment based on a small-scale problem solvable for the optimal control, this research investigates the effect of the re-solving frequency for the popular bid-price control, which is based on a static linear programming formulation. In addition, the control method based on the parameterized function is tested in the simulation experiment, as it does not involve problem-resolving and can serve as the basis for evaluating the re-solving frequency. Based on the results in the simulation experiment, the method based on the parameterized function can achieve the performance very close to the optimal control. As for the bid-price control, a high re-solving frequency is needed for a similar result, and the revenue is significantly reduced if the bid prices are not updated frequently.