Академический Документы
Профессиональный Документы
Культура Документы
*Humboldt-University Berlin, Institute of Mathematics, Unter den Linden 6, Berlin, 10099, Germany
Tel: +49 30 2093 2353; Fax: +49 30 2093 2232; E-mail: romisch@mathematik.hu-berlin.de
**Lufthansa Systems Berlin GmbH, Fritschstrasse 27–28, Berlin, 10585, Germany
Tel: +49 30 3400 7178; Fax: +49 30 3400 7100; E-mail: klaus.weber@lhsystems.com
Page 265
A new approach to O&D revenue management
Page 266
Möller, Römisch and Weber
In the next section, the stochastic pro- To state the stochastic programming
gramming model for O&D revenue man- (SP) model, it is assumed that S scenarios
agement is established in scenario and node with probabilities ps>0, s=1, . . . , S, of
formulation. Furthermore, the generation the booking demand and cancellations pro-
of a booking and cancellation scenario tree cess are given. These scenarios may be
from individual scenarios is described. In obtained from stochastic demand models
the final sections, preliminary numerical and by relying on expert knowledge,
experience is reported, and concluding respectively.
comments are given.
Scenario-based SP model
STOCHASTIC PROGRAMMING MODEL To set up the SP model, some further
notation is needed. We denote the index
Modelling set of itineraries containing leg l (ie, the
An O&D network is considered, consisting incidence set) by I l ( {1, . . . , I}, the
of I origin–destination itineraries, J fare number of compartments on leg l by M(l)
classes, K points of sale, L legs with M(l) and the index set of fare classes of compart-
compartments in each leg l = 1, . . . , L. ment m on leg l by Jm(l) ( {1, . . . , J}.
Let the booking horizon be subdivided Further input data are the fares fi,j,k,t and
into T booking subintervals with data col- the capacities Cl,m of compartments m [ {1,
lection points (dcps) t = 0, . . . , T. The . . . , M(l)} and legs l.
booking process is controlled over time by The stochastic input variables are the
decisions on protection levels Pi,j,k,t for booking demand dsi,j,k,t and and the cancel-
each fare class j [ {1, . . . , J}, itinerary lation rates gsi,j,k,t. The bookings bsi,j,k,t and
i [ {1, . . . , I}, point of sale k [ {1, . . . , K} the cumulative bookings Bsi,j,k,t represent
and at each dcp t = 0, . . . , T – 1. The the stochastic state variables of the model
decisions at t are made for the next book- while the protection levels Psi,j,k,t are the
ing interval (t, t + 1] based on the pre- stochastic decisions. Here, the superscript s
vious process of bookings and cancellations always refers to scenario s. The expected
up to t and recursively over time. Protec- total revenue is considered the objective
tion levels are upper bounds for the inven- function, where total refers to the whole
tory of booked, uncancelled seats. O&D network and booking horizon, all
It is assumed that the fares and the com- fare classes and points of sale.
partment capacities are given, ie, they are Summarising, our scenario-based sto-
deterministic input variables. The booking chastic programming model consists in
demand and the cancellation processes are maximising
regarded as a multivariate stochastic process
T
{xt}t=0 over time, where x0 represents a X
S X
T X
I X
J X
K
known deterministic starting value. The s fi;j;k;t ðbsi;j;k;t csi;j;k;t Þ
components of the random input vector xt s¼1 t¼1 i¼1 j¼1 k¼1
at t are the stochastic booking demands ð1Þ
di,j,k,t and stochastic cancellation rates ci,j,k,t.
s
Hence, xt is a 2IJK-dimensional random subject to all protection levels Pi;j;k;t satisfy-
vector whose components are statistically ing
dependent and, furthermore, the random
input vector xt depends on its history (x0, s
ð1 i;j;k;t ÞBsi;j;k;t Pi;j;k;t1
s
ð2Þ
x1, . . . , xt–1).
Page 267
A new approach to O&D revenue management
where Bsi;j;k;t are the cumulative bookings, cancelled: A ticket may have been given
ie, back or the passenger may have been
rebooked to another flight. In order to
Bsi;j;k;0 :¼ 0i;j;k ;
B uncover cancellation fees or refunds cor-
Bsi;j;k;t :¼ BSi;j;k;t1 þ bsi;j;k;t ð3Þ rectly, the original tariff information
should be available. Even in passenger
with bsi;j;k;t satisfying the demand constraints name records, however, often only book-
bsi;j;k;t di;j;k;t
s
ð4Þ ing class information is stored. Approxima-
tions could be received by analysis of
and the leg capacity limits coupon information from check-in and
PNR data, but most airlines have not
X X X
K
s
Pi;j;k;T1 Cl;m ð5Þ established such process.
i2I l j2J m ðlÞ k¼1
Equation (3) describes the update of the
cumulative bookings starting with the
For some # 2 ð0:0; 0:5 the cancellations are initial value B 0i;j;k 2 Z. The constraint (5)
approximated by expresses that, for each leg, the correspond-
s ing protection levels may not exceed the
i;j;k;t Bsi;j;k;t i;j;k;t1
s
Bsi;j;k;t1 #
physical capacities of the compartments on
csi;j;k;t < ð6Þ the day of departure. The latter implies no-
s
i;j;k;t Bsi;j;k;t s
i;j;k;t1 Bsi;j;k;t1 þ# show based overbooking is not part of the
model. This is another simplification,
Finally, the integrality and non-negativity owing to the study character of the work.
constraints Since constraint (5) applies at time of depar-
ture only, overbooking is possible during
bsi;j;k;t ; csi;j;k;t ; Pi;j;k;t
s
2Z ð7Þ
the entire booking period. Actually, it is
more demanding to model overbooking to
bsi;j;k;t ; csi;j;k;t ; Pi;j;k;t
s
0 ð8Þ compensate cancellations than to model
overbooking to counteract no-shows.
and the non-anticipativity constraints have
While the protection levels, the number
to be satisfied, the latter meaning that
of bookings and the number of cancella-
decisions at t only depend on the data tions have to be non-negative integers by
until t. (9) nature, the constraint (9) expresses how the
information flow evolves over time. (9)
Here, (1) corresponds to the total expected may be modelled by linear equations in
revenue. Like the revenue values processed various ways, see Ruszczynski and Shapiro
in real revenue management systems, the (2003, Chapter 3.6) and Römisch and
fares in the model are divided according to Schultz (2001). Altogether, the model (1)–
booking period, itinerary (and thus origin, (9) represents a large scale multistage sto-
destination, and flight period), booking chastic integer program.
class, and point of sale. However, the objec-
tive function is simplified, as full refund of Input scenario trees
cancelled tickets is assumed. This is justifi- The non-anticipativity constraint (9)
able, because it is not a mathematical pro- implies that the finitely many scenarios
blem to take cancellation fees and refunds fts gT
t¼0 ; s ¼ 1; . . . ; S, can be represented in
into account, but a practical one. In fact, it is the form of a scenario tree. The scenario
quite difficult to determine from an airline’s tree is based on a finite set
database for what reason a booking was N ¼ f0; 1; . . . ; Ng of nodes that are
Page 268
Möller, Römisch and Weber
P
arranged at the stages t = 0, . . . , T . The that n2N t n ¼ 1 and tðnÞ ¼ fn gn2N t , for
root node n = 0 is the only node at stage each t = 0, 1, . . . , T.
t = 0. The number of nodes at stage t = 1 The generation of scenario trees that
corresponds to the number of different rea- approximate the stochastic input process
lisations of 1 . Each of these nodes is con- ft gT t¼0 is a challenging task when solving
nected with the root node by an arc. In multistage stochastic programs. In Dupa-
general, each node n 2 N ; n 6¼ 0 has a c̆ová et al. (2000), an overview of scenario
unique predecessor node denoted by n– and tree generation techniques is provided.
a set N þ ðnÞ of successor nodes. Each node More recent contributions are based on the
in N þ ðnÞ is connected with n by an arc. moment-matching principle (Høyland and
The set {0, . . . , n–, n} of recursive prede- Wallace, 2001), the use of distances of
cessors of n is denoted by path(n), which probability distributions (Pflug, 2001) and
refers to the path from the root to n. t(n) (Gröwe-Kuska et al., 2003), and Quasi-
denotes the number of elements in path(n) Monte Carlo methods (Pennanen, 2004),
minus 1 and, thus, refers to the stage to respectively.
which n is arranged, ie, the nodes in Next, the scenario tree construction
N t :¼ fn 2 N : t ¼ tðnÞg correspond to approach presented in Gröwe-Kuska et al.
the different realisations of t . Nodes n (2003) is briefly described. It assumes that
belonging to set N T have the property a finite number of individual scenarios
N þ ðnÞ 6¼ 1 and are called leaves. Hence, a fts gTt¼0 with probabilities ps > 0, s = 1,
scenario corresponds to a path from the . . . , S, and common root node, ie,
root to some leaf, ie to path(n) for some 01 ¼ . . . ¼ 0S is given. These scenarios
n 2 N T , and its probability is renamed by may be obtained from simulations of a
pn. We also say that pn is the probability of parametric statistical model (eg based on
the leaf n. Clearly, we have time series analysis) or from a non-
fn gn2N T ¼ fs gSs¼1 . The probabilities of parametric model (eg by resampling
nodes P n2= N T compute by the recursion methods). This fan of individual scenarios
n :¼ nþ 2N þ ðnÞ nþ . Clearly, we have is modified by a procedure of recursive
bundling and deletion of similar scenarios,
respectively, leading to a tree structure.
Its methodology is based on the scenario
reduction techniques developed by Dupa-
Figure 1: Scenario tree with T = 4, N = 21 and c̆ová et al. (2003) and Heitsch and
11 leaves Römisch (2003) and employs these techni-
ques backwards in time starting at t = T.
The bundling and deletion process relies
on computing and bounding the Kantoro-
vich distance of the original probability
distribution DðÞ given by the individual
scenarios and their weights and of the dis-
tributions of the approximate trees. If
f~t gTt¼0 and q 0; ¼ 1; . . . ; S; denote
the scenarios and weights of another dis-
crete probability distribution D(x~), the
Kantorovich distance of D(x) and D(x~) is
given by
t¼0 t¼1 tðnÞ T
Page 269
A new approach to O&D revenue management
ðDðÞ; Dð~ÞÞ :¼ X X
Ttþ1
Page 270
Möller, Römisch and Weber
After three reduction and bundling steps n fi;j;k;tðnÞ ðbni;j;k cni;j;k Þ ð12Þ
n¼1 i¼1 j¼1 k¼1
at t = 3, 2 and 1, the final result is
shown in (d). The final scenario tree exhi- n
subject to all protection levels Pi;j;k satisfy-
bits a possibly different branching struc- ing
ture at all stages, which is detected by the
n n
algorithm. ð1 i;j;k ÞBni;j;k Pi;j;k ð13Þ
Page 271
A new approach to O&D revenue management
where Bni;j;k are the cumulative bookings, ie and process protection levels (or booking
protects) on leg, booking class level or bid
B0i;j;k :¼ B
0i;j;k prices on leg, compartment level. Thus,
Bni;j;k :¼ Bni;j;k
þ bni;j;k ð14Þ processing of the protection levels in the
present approach requires seamless booking
with bni;j;k satisfying the demand constraints control which is actually practised for par-
tial or complete flight networks at some
bni;j;k di;j;k
n
ð15Þ airlines. In the case of seamless control
booking requests are not responded to by
and for all n 2 N T1 the leg capacity limits the computer reservation system (CRS)
X X X
K but processed by the airline’s inventory
n
Pi;j;k Cl;m ð16Þ system or a separate ‘availability processor’
i2I t j2J m ðlÞ k¼1 (AP). In the following, use of an AP is
assumed.
For some # 2 (0.0, 0.5] the cancellations
The protection levels as solutions of the
are approximated by
multistage stochastic programs form a
n n n
i;j;k Bni;j;k i;j;k Bi;j;k # (multivariate) stochastic process over time
cni;j;k < ð17Þ with the same structure as the input data.
n n n
This differs from methods which calculate
i;j;k Bni;j;k i;j;k Bi;j;k þ# protection levels for the entire remaining
booking period, but is similar to dynamic
Finally, we have the non-negative integer
program approaches, which compute bid
constraints
price vectors for all dcps to come. For
bni;j;k ; cni;j;k ; Pi;j;k
n
2 Z ð18Þ practical operation, protection levels may
be operated as follows.
bni;j;k ; cni;j;k ; Pi;j;k
n
0 ð19Þ
— The (deterministic) protection levels of
while the nonanticipativity constraints are dcp t = 0 may be taken directly to the
satisfied by construction. Altogether, the AP. At dcps t ¼ t0 ; t0 2 f1; . . . ; T 1g,
model (12)–(19) represents a large-scale the further process depends on the
structured integer program which is of actual booking inventory Bi;j;k;t0 :
smaller dimension compared to its scenario — If there exists a node n such that t(n) =
formulation. More precisely, it contains t0 and Bi;j;k;t0 ¼ Bni;j;k , the protection
n
IJK+4IJK(N–1–S)+3IJKS
PL variables and levels Pi;j;k can be uploaded to the AP
3IJK(N–1) + ( l¼1 M(l))S constraints. for controlling the booking process
Since all variables are non-negative and until the next dcp.
the bookings are bounded from above, the — If Bi;j;k;t0 6¼ Bni;j;k for all nodes n with t(n)
objective function is also bounded from = t0, the stochastic optimisation model
above. Hence, the LP relaxation of the is restarted with a new input scenario
integer program, ie when the constraint tree having its root node at t0.
(18) is ignored, is solvable. For its solution, — If, for any reason, such a re-optimisa-
any standard LP solver may be used. tion is not possible, then information
Owing to the distinction of itinerary, on the probability distribution (means,
booking class, and point of sale the protec- quantiles etc) of the relevant protection
n
tion levels ðPi;j;k Þn2N allow considerably levels (determined by the difference
accurate control of booking requests. between Bi;j;k;t0 and fBni;j;k gtðnÞ¼t0 ) could
Today’s inventory systems usually hold be taken to compute a fallback solution.
Page 272
Möller, Römisch and Weber
L I J K T M(1)
1 1 14 1 18 3
S N No. of No. of
variables constraints
Page 273
A new approach to O&D revenue management
Figure 5: Cumulative passenger demand and protection level for selected fare classes
Page 274
Möller, Römisch and Weber
Page 275
A new approach to O&D revenue management
Page 276