以解構方式求解航空網路營收管理問題之競價控制法則

Abstract

自1979年解除航空管制後，眾多航空公司應用營收管理之技巧，並發展了複雜且多樣化之各式費率艙等來提高收益，一方面也滿足乘客需求。 航空網路的艙位存貨管理即是針對不同起訖點與費率需求控管各航段座位。然而，網路包含了多個航段，較單一起訖點問題複雜得多。跨航段的起訖點需求牽涉到資源的共享，所以在決定最適艙位配置時，若只針對單一個航段來考量並無法保證能使整個網路之總收益最大。 近幾年來，解決航空網路營收管理之問題最成功的方法之一為「競價法」，所有競價值是每個航段用來訂定控管決策的一個門檻。根據此門檻值，若該艙等之費率大於門檻值，則接受此訂位。反之，則拒絕此訂位，而跨多個航段的競價值則是各個航段線性之累加。 訂定競價門檻值是一個相當困難的工作。本研究發展了一個以解構方式為基礎之演算法來找出每個航段之競價門檻值處理航空網路營收管理之問題。相關之數值測試，係以可以運用動態規劃求得最佳解之小型營收管理問題為基礎，應用所求得知競價門檻值最為控管決策，並進而將球得知營收與最佳值進行比較。除此之外，並以無機位控管的先到先服務方式進一步評估此演算法之優劣。

Since airline deregulation in 1979, many airlines began to apply revenue management (RM) techniques. They have developed very complex and diverse fare classes to increase the revenue and, at the same time, to satisfy the demand of passengers. Seat inventory control for air network is the practice to allocate the seats of the flight legs among different passenger itineraries and fare classes. Nevertheless, the network problem with multiple legs is more complicated than the single-leg version. As an itinerary can involve to multiple resources, when making the seat inventory control decisions, focusing on an individual flight leg does not guarantee that the total revenue will be maximize across the entire network. In recent years, one of the most successful approaches for solving the airline network RM problem is the bid price control, in which the bid prices of the legs are used as a threshold to derive the accept-or-reject decisions. Based on the bid prices, the resulting policy accepts a booking request only if its fare is greater than the sum of the bid prices of the associated itinerary. The most difficult part of this approach is to determine the bid prices. We develop a decomposition-based algorithm of the airline network RM problem to find the bid price of each flight leg. As for the numerical experiment, the algorithm is applied to several small-scale network RM problems, for which the optimal solution can be found by dynamic programming. The expected revenues based on the bid price control are compared with the optimal values. In addition, the case of first-come-first-serve is tested to further evaluate the solution quality of this heuristic algorithm.