航空公司超額訂位控制策略之研究

Abstract

本研究乃運用一次決策與最佳邊界（Optimal boundary）的概念，構建單一費率、及兩費率等級的超額訂位模式。針對下列五項與超額訂位相關議題，作進一步深入的探究：1.超賣機位賠償成本之設定。2.訂位上限之設定（彈性艙位配置）。3.訂位型態與艙位配置之關係。4.隨機需求下，期望總營收之求解。5.複雜連續之訂位控制策略與一次決策之適用性。 本研究經由對上述議題之探討發現： 1.不論是線性或非線性超賣賠償成本函數的設定對超額訂位問題皆有重要的影響，考慮非線性超賣賠償成本函數所求算出的訂位邊界明顯小於考慮線性超賣賠償成本函數所求算出的訂位邊界（在相同的超賣賠償成本參數設定下）。 2.單一費率等級訂位上限之決定與旅客訂位需求的機率分配函數無關，只與旅客出現率、超賣賠償成本、及艙位容量有關。 3.在兩費率等級之超額訂位問題中，低費率等級訂位上限的訂定與高費率等級的旅客訂位需求機率函數有關。在其他條件不變之下，若高費率等級的旅客訂位需求的變異數越大，則低費率等級的訂位上限越高；反之，若高費率等級的旅客訂位需求變異越小，則低費率等級的訂位上限越低。 4.提出『期望邊際訂位營收』此一新名詞。在不考慮取消訂位與未出現下，『期望邊際訂位營收』即為期望邊際艙位營收，此時期望邊際艙位營收為艙位數之非遞增函數。 5.在考慮旅客會取消訂位與未出現下，期望邊際艙位營收之求算若不加入旅客之訂位需求（亦即隱含著旅客訂位需求量高於訂位上限），期望邊際艙位營收為艙位數之非遞減函數。然而，若加入旅客之訂位需求，則期望邊際艙位營收先為艙位數之非遞減函數，隨著艙位容量增大而為艙位數之非遞增函數。因為當艙位容量較大時，旅客之訂位需求不會總是大到足夠填滿所有艙位。 6.若在訂位過程中旅客訂位動態（訂位到達、與取消訂位）與以往同質性班次之訂位資料沒有太多差異，則不斷地重複檢查程序對訂位控制應不會有顯著影響。除非我們能非常確定連續檢查程序的調整方法對營收有相當的影響效果，且航空公司有足夠的資訊（資料）作訂位控制之調整，否則連續決策將不見得更有效。

In this study, an optimal boundary concept is developed for one-time decision airline overbooking problem in the cases of single-fare class and two-fare class. Five issues including the impact of oversale penalty, the establishment of booking limit, the relationship between pattern and seats allocation, the expected total revenue under stochastic demand, and the applicability of sophisticated continuous booking control strategy and one-time decision are specially addressed. The research has brought out the following results: 1.Both linear and non-linear oversale penalty cost functions have great impact on overbooking. The booking boundary computed from non-linear oversale penalty function is smaller than that from linear one when the oversale penalty cost parameters are set the same. 2.The optimal overbooking limit in the case of single fare has nothing to do with the booking demand. It only relates to passenger's show-up rate, oversale penalty cost, and the flight capacity. 3.In the case of two-fare, the booking limit for low fare is related to high fare booking demand probability function. When the other conditions remain unchanged, the higher the demand variance of high fare is, the higher low fare booking limit will be. 4.The new term "expected marginal booking revenue" is proposed. It equals to the expected marginal seat revenue only when cancellation and no-show are ignored. The expected marginal seat revenue is a non-increasing function of seats. 5.When booking demand distribution function is ignored (It implies that the booking request is always higher than the booking limit) and cancellation and no-shows are considered, the expected marginal seat revenue is a non-decreasing function of seats. However, if passengers' demand distribution function is incorporated, the expected marginal seat revenue initially will be a non-decreasing function then followed by a non-increasing function of seats. Since the booking request is not always large enough to fill up all the seats as capacity becomes large enough. 6.The continuous review process will not have significant impacts if passenger's booking pattern is not deviate far away from the one used in the control process which, in general, is generated from the historic flights' booking information. That is, the continuous review process will not be efficient unless we are very sure that information is sufficient enough to make adjustments and better decisions on booking control.