A Web Integration Framework for Cheap Flight Fares
Manuel Sánchez
1
, Juncal Gutiérrez-Artacho
1
and Jorge Bernardino
1,3
1
Superior Institute of Engineering of Coimbra, Polytechnic of Coimbra, Coimbra, Portugal
2
Department of Translation and Interpreting, University of Granada, Granada, Spain
3
Centre of Informatics and Systems of the University of Coimbra, University of Coimbra, Coimbra, Portugal
Keywords: Data Integration, Web Services, Relational Database Systems, Graph Systems, Big Data, Data Mining.
Abstract: Travel agencies offer their services via the Internet, which creates new methods of communication and
connection between customers and third companies. Due to the difficulty that the management of large
volume of flight routes represents, it is necessary to capture the information provided by airlines through a
variety of services, providing end customers with competitive fares. In this paper, we analyse the
information source of flight fares offered by airlines, studying the difficulties, limitations and costs involved
in accessing these data. A framework that explores the possibilities of finding "hidden" flight fares that
result in much cheaper options in comparison to the average price of each flight route will be presented.
1 INTRODUCTION
Travel agencies have been around for decades, but
throughout their history, and technological progress,
they have been evolving and adapting to a new
environment: the Internet. In order to make the
transition to the Cloud, travel agencies have been
transformed into a new kind of service provider,
known as OTAs (Online Travel Agencies). This new
business model, which offers its services over the
Internet, has forced companies to create new
methods of communication and connection between
customers and third companies, through the use of
Web Services.
A Web Service is a technology that uses a set of
protocols and standards that are employed to
exchange data between different applications on
different platforms, using the Internet to transmit
these data. Web Services are supplied on a uniform
programming interface called an API (Application
Programming Interface) (Binstock, 2015).
Online Travel Agencies provide Programming
Interfaces to extend their services to third parties,
and this is the key for accessing relevant information
on airfares in order to enable automated
management processes and the optimization of flight
searches. In the field of OTAs, we focus our work
on the flight management segment, and airline ticket
acquisition services.
Due to the size represented by the treatment of
this large volume of data (flight routes on a global
scale), it is necessary to address the relevant topic of
Big Data (IBM Big Data, 2015) in terms of the
capture of information provided by the airlines
through a variety of services and products
(Guterman, 2015). In this paper, we propose a
solution for processing and analyzing all obtained
information, to achieve favourable patterns for
future searches and thus providing customers with
the best fares.
The motivation behind this work is to present
and analyze business and technical tool options on
the market, and develop a new feature increasingly
demanded among customers. With our framework
we find the best deals on airfares and with them
make combinations that result in new flight routes,
offering customers the possibility of travelling to
places they would not have considered. As far as we
know, this new feature is missing from all of these
companies’ services. Despite this, it is partially
achievable, but in order to do so it is necessary to
perform an exhaustive manual search to find a trip
with a truly affordable price.
In practice, the main contributions of this work
are:
• To analyse the tools provided by Online Travel
Agencies (OTAs) and major flight search engines
• To offer their services through third companies,
which will allow the development of new features.
260
Sánchez, M., Gutiérrez-Artacho, J. and Bernardino, J.
A Web Integration Framework for Cheap Flight Fares.
DOI: 10.5220/0006287702600267
In Proceedings of the 13th International Conference on Web Information Systems and Technologies (WEBIST 2017), pages 260-267
ISBN: 978-989-758-246-2
Copyright © 2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
• To obtain the best possible fare for a particular
route
• To discuss why current systems do not
implement this feature automatically.
The remainder of this paper is organized as
follows. Section 2 gives an introduction to the
current systems available on the market, what are the
best options, when thinking about using these
services and what they do not provide to their
customers. Section 3 analyses two of the most
widespread classes of Database Management
Systems (DBMS) on the market, their functions and
features, and the choice of a tool according to the
nature of the problem. Section 4 explains the
problems encountered in establishing a single
information source for flight fares. Section 5
describes our proposed web integration framework
explaining the architecture of the system developed
and giving an example of the results of flight fares
using our solution. Finally, in Section 6 we present
the conclusions and indicate future work.
2 BACKGROUND
Nowadays there are services that OTAs increasingly
use to grasp the attention of potential customers,
which include attractive features that enable greater
accessibility and dynamism. Many of these
companies are well known for offering the best deals
on airfares, others for having flexible tools to specify
a date for travelling, and others also offer the
possibility of choosing multiple destinations.
The main difficulty that these OTAs have when
offering possible flight combinations are the
restrictions applied to tariffs by the airlines
companies. These difficulties differ from temporary
restrictions on advance ticket purchase, up to
combination restrictions between different flights,
which can become quite complex, making too
difficult processing the information for determining
all possible combinations. An example of a
temporary restriction could be the application of a
particular fare that can only be carried out with the
purchase of a ticket 14 days before departure (Valles
& VanLoy, 1991).
We can see an example of combinations of
airfare restrictions when attempting to combine this
fare with another provided by a different company
or airline alliance. This results in an incompatible
approach to applying these prices under a unified
buying process, because the offers are only valid
when combined with other flights from the same
company.
Due to the complexity of the restrictions on the
airfares, the implementation of an upper layer of
analysis of these results is necessary, which allows
us to achieve optimal results searching fares
independently, without infringing the restrictions
applied to each one.
Some examples of implemented systems that
provide similar services with fare graphic
visualizations are companies like Vayant, Skypicker
and Kayak. These companies have systems that,
starting from a predetermined origin, find flight
routes with the existing lowest price across all dates,
so that if we choose a point of origin they can find
the lowest fare flight for each of the possible
destinations.
Other OTAs such as Skyscanner offer total
flexibility for choosing travel dates, being able to
search a range of weeks, months, or even the full
year, thus obtaining the best possible price for a
given route. Figure 1 shows an example of the
Skyscanner application.
Figure 1: Skyscanner application.
Skyscanner is a complete platform where we can
search with flexible dates and general locations such
as a city which may have several airports, or an
entire country, although it is not possible to suggest
the best deals to the client on routes not established
by the user, and they cannot make combinations that
result in new flight routes. A real example of this
can be observed in Figure 2, where we found a flight
from Cancun (CUN, Cancun, Mexico) to Brussels
(BRU, Brussels, Belgium) for €80, which is an
excellent deal, much cheaper than the next lowest
price, €1,000. If we had not made a search between
Mexico and Belgium we would never have found
this offer, because Skyscanner does not display these
suggestions automatically.
3 DATABASE MANAGEMENT
TOOLS
In this section we analyse and compare the most
A Web Integration Framework for Cheap Flight Fares
261
Figure 2: Great fare from Cancun to Brussels.
relevant information storage systems in the market:
Relational Databases and Graph Databases. These
systems are tested considering two metrics: data
storage size and performance using typical queries.
We also studied the services and facilities offered by
each of these databases systems. At the end of the
section we discuss the nature of the problem, i.e.,
possible combinations of routes based on nodes and
relationships to get from an origin point to a given
destination point.
3.1 Relational Databases
Relational Databases have been the standard system
for storing and accessing information used by most
systems since the early '80s, and even today. Most
current business models, due to their characteristics,
internal processes and restrictions, are naturally
adapted for this architecture based on ACID
(Atomicity, Consistency, Isolation and Durability)
properties, helping to understanding and integrate
traditional models.
An important goal of any database system is to
model the real world accurately in a consistent
manner with the user's perception of the data.
Unfortunately most traditional DBMS do not
provide adequate integrity features to ensure the
accuracy of data in their databases. Correspondingly,
most of traditional DBMS do not provide the
necessary optimally modelling tools. They present
difficulties because these databases are more
oriented to define the characteristics that identify the
relationships of the problems that determine how
they relate to each other.
The advantages provided by a relational DBMS
(RDBMS) are based on the capacity for integration
and standardization of the language and architecture,
serving a homogeneous model on mathematical
rules. We also noticed the disk storage a requirement
of information, when compared to other DBMS is an
important feature to consider. In Table 1 we show
the experimental results of the volume of data
occupied in disk with structures of random integers,
and strings with 8KB, and 32Kb. For these
experiments we choose MySQL as a tool for
Relational Databases, and Neo4j for Graph
Database.
Table 1: Size comparison of different data types.
Database #Nodes
Data
Type
MySQL
Size
Neo4j Size
1000int 1000 Int 0.232M 0.428M
5000int 5000 Int 0.828M 1.7M
10000int 10000 Int 1.6M 3.2M
100000int 100000 Int 15M 31M
1000char8k 1000 8K Char 18M 33M
5000char8k 5000 8K Char 87M 146M
10000char8k 10000 8K Char 173M 292M
100000char8k 100000 8K Char 1700M 2900M
1000char32k 1000 32K Char 70M 85M
5000char32k 5000 32K Char 504M 406M
10000char32k 10000 32K Char 778M 810M
100000char32k 100000 32K Char 6200M 7900M
As we can see from Table 1, the results of
storing data into Neo4j database takes between 1.25
to 2 times the size of the MySQL database. One
disadvantage of using relational databases to tackle
this issue is at the same time its best feature, strong
constraints of the model tables/relationships and
structured language (SQL) that needs to be applied.
Therefore discards using an architecture of this type,
opting for the use of other NoSQL models, namely
Graph Systems (Kamel, 1994). Systems based in
NoSQL allow dynamic structures, and are faster to
process the information, a key to dealing with
massive data volume as will see in the next section.
3.2 Graph Databases
Graph Databases are based on NoSQL model that
have gained recognition in recent years due to new
features provided when modelling a database. The
term “NoSQL”, as a term for modern web data
WEBIST 2017 - 13th International Conference on Web Information Systems and Technologies
262
stores, first began to gain popularity in early 2009. It
is a topic that has gained credit from the IT
community but has yet to garner large-scale
academic study.
The advantages provided by these systems are
evident; the freedom to organize the structures so
that they are not restrictive, and the speed to process
information stored (higher than RDBMS when
larger is the amount of information to be processed
in a query). Table 2 shows the queries to simulate
some of the types of queries used in our provenance
systems. For example, traversals are necessary to
determine data objects (nodes) derived from or
affected by some starting object or node:
Q0: Find all orphan nodes. That is, find all
nodes in the graph that are singletons, with
no incoming edges and no outgoing edges.
Q4: Traverse the graph to a depth of 4 and
count the number of nodes reachable.
Q128: Traverse the graph to a depth of 128
and count the number of nodes reachable.
Table 2: Structural query results (in milliseconds).
Database
MySQL
Q0
Neo4j
Q0
MySQL
Q4
Neo4j
Q4
MySQL
Q128
Neo4j
Q128
1000int 1.5 9.6 38.9 2.8 80.4 15.5
5000int 7.4 10.6 14.3 1.4 97.3 30.5
10000int 14.8 23.5 10.5 0.5 75.5 12.5
100000int 187.1 161.8 6.8 2.4 69.8 18.0
1000char8K 1.1 1.1 1.1 0.1 21.4 1.3
5000char8K 7.6 7.5 1.0 0.1 34.8 1.9
10000char8K 14.9 14.6 1.1 0.6 37.4 4.3
100000char8K 187.1 146.8 1.1 6.5 40.9 13.5
1000char32K 1.3 1.0 1.0 0.1 12.5 0.5
5000char32K 7.6 7.5 2.1 0.5 29.0 1.6
10000char32K 15.1 15.5 1.1 0.8 38.1 2.5
100000char32K 183.4 170.0 6.8 4.4 39.8 8.1
For the traversal queries, Q0, Q4, and Q128, Neo4j
was clearly faster, sometimes by a factor of 10, as
detailed in Table 2. This was expected since
relational databases are not designed to do
traversals.
We conclude that the Graph Database systems
are the most optimal for searching existing
connection between nodes, providing response times
much lower than those obtained using Relational
Databases, which is the critical problem of flight
management systems.
3.3 Graph Database Tool
Based on previous experiments, we have chosen the
Graph DBMS architecture. In this section, we
discuss which is the most suitable tool to implement
a system of this type, which contains several
requirements when implementing this solution.
These requirements are the speed of data processing,
heterogeneity of access to information, migration of
the platform, or a possible scalability of the system.
Among all the possibilities, and based on a study of
efficiency of Graph DBMS platforms, we have
determined that the most efficient tool could be
Neo4j (Woodie, 2015). It is a Graph DBMS that we
can get in free community versions, or Enterprise,
which involves an economic outlay, taking
advantage of technical assistance, which can be of
great help in certain occasions.
After choosing the best Graph DBMS tool used to
develop the application, it is necessary to analyze in
which platform will deploy it. When choosing where
to host the database, there are several possibilities:
hosting on AWS / EC2, Windows Azure, or Cloud
Hosting providers like GrapheneDB, GraphStory,
Structr, etc. Thinking about the flexibility of the
system, the best option may be to acquire a server in
a hosting. This option is based in being able to
deploy databases, implement Data Mining system,
and develop diverse features under the same hosting.
3.4 Problem Definition
When thinking about how to approach the solution
to the problem of connections between different
flights, we have the idea of a large interconnected
network in which all points will not be
interconnected, but we have a variety of possibilities
to arrive at the desired point from a given origin. By
this way we can model the problem, and the best
option is using graphs, because the essential
information resides in the interaction between
connections, naturally expressed by graphs DBMS.
An example of the nature of the problem is shown
in Figure 5. In order to determine the possibilities of
reaching from Paris-Orly Airport (ORY, Paris,
France) to the International Brussels Airport (BRU,
Brussels, Belgium), we have not defined a direct
A Web Integration Framework for Cheap Flight Fares
263
route, but if we can determine different possibilities
to reach final destination. The proposed objective is
to reach any destination at the lowest possible cost.
In this case, the intermediate routes that could be
selected to reach the desired destination could be
through the connection at Barcelona airport (BCN,
Barcelona, Spain), Barajas airport (MAD, Madrid,
Spain) and Porto airport (OPO, Porto, Portugal). The
result of final route would be determined by the
minimum possible path and the lowest price found
among all these possibilities.
Figure 3: Routes between airports example.
4 INFORMATION ACCESS
POINT PROBLEMS
The main way to access airlines flight fares are
through GDS (Global Distribution System). GDS is
a network operated by a company that enables
automated transactions between third parties and
booking agents in order to provide travel-related
services to end consumers. A GDS can link services,
rates and bookings, consolidating products and
services across all three-travel sectors (book flight,
book hotel, car hire).
We can find two variants of the GDS systems
(Jonas, 2013); one of them is GNE (Global
Distribution System New Entrant). It does not
include incentives in transactions like regular GDSs,
therefore the profit margin for travel agents is less,
and will offer a more affordable price for the
purchase of flights.
The LCC (Low Cost Carrier) is another variant
that includes the largest listing group flights of low
cost companies and also offering their offers on their
own websites. Therefore, it is also a good option for
our use.
A GDS system is composed of several modules
management programs called CRS (Computer
Reservation System), which is a computerized
system used to store and retrieve information and
conduct transactions related to air travel, hotels, car
rental, or activities. Originally designed and
operated by airlines, CRSs were later extended for
the use of travel agencies (AXSES, 2013).
The main advantage of using these systems is the
centralization of the flights of different airlines,
which provides the same source information point,
and with the same interface for access to data
(Hospitality Net, 2015).
The major disadvantage of the centralization of
these data is the free use of them, which becomes
rather difficult to handle for a single user or a small
business, due to the access and use of these systems
includes a payment. The payment is normally under
a SaaS license, whose cost is determined by the
amount of transactions that are made to the system
(Consumer Reports WEB Watch, 2015).
The GDS systems are not the source of the
information of the flights from airlines. On the
contrary, there are two main agencies that saved and
updated every hour, virtually in real-time
information provided by the airlines. One of them is
ATPCO (Airline Tariff Publishing Company), and
the other is SITA (Société Internationale de
Télécommunications Aéronautiques).
ATPCO is a corporation that publishes the latest
airfares for more than 500 airlines, multiple times
per day, and provides fare data in an electronic
format, which make the information suitable for
computer processing. The only competitor to
ATPCO is SITA, who distributes some fares only in
Asia, Africa and Europe.
Fares are distributed hourly each day and airlines
carefully monitor new public fares filed by their
competition for publication through their systems.
Once time these corporations have distributed the
fares, airlines detect the action of other airlines
increasing or decreasing their fares for specific
connections, and then use this information to set
their own pricing strategy. For instance, if they see a
competitor introducing special promotional pricing
between two cities, they may want to quickly react
by filing their own special fares.
Access to the data provided by these major
corporations is very expensive and imposes several
restrictions, such as having an accredited title of
travel agent, which is obtained by performing a test,
and paid substantial fees. Due to these difficulties,
the best option would be to access the information
by consuming Web services that provide companies
more focused to the end customer, the OTAs (Online
Travel Agencies).
WEBIST 2017 - 13th International Conference on Web Information Systems and Technologies
264
5 WEB INTEGRATION
FRAMEWORK
In this section we explain our web integration
framework, describing its architecture and also
giving a working example of our proposal.
5.1 Proposed Architecture
The modelling of the solution has been made from
the information source that provides Skyscanner
through their offered API. The information
processing consist in four fundamental steps, and
assume a starting point in the system where the
database has stored the list of existing airports and
all the routes that connect it directly (see Figure 6).
To dispose the system at this early stage, it is
necessary to extract the relevant information from
airports and routes from another sources, in this case
it will be FlightStats and OpenFlights. We can
access the list of active civil airports through these
two organizations, and active routes between each of
them through OpenFlights. To obtain the
information through OpenFlights is only necessary
to download the appropriate files.
If we make requests to FlightStats, the response
will be in JSON format, and we must transform and
store the data in our database. When storing airports,
we get as a result, nodes that will serve as a
connection between possible routes in our system.
Once we have in the system the information about
airports and routes, we can make requests in a
logical and structured way through the Skyscanner
API to obtain flight offers. Skyscanner API is a
useful and essential tool to develop this solution; it
obtains the "best offer" for a given route by a simple
request (offers are subject to airfare restrictions). To
determine the extent of an offer, it is necessary to
know the average price that has a route, and in this
case, Skyscanner also offers an optimal solution.
Requests are performed following the scheme set
out in our database for all stored routes of available
airports. With this requests, we complete a first layer
of flight fares, and the average prices for each route.
This process will be processed with an algorithm
of minimum nodes cover (airports), to avoid
overloading the system with duplicate requests.
At this stage, we already could publish flight
offers, because fares are the lowest found searching
by default. It is also advisable to develop a parallel
system of Offer Management, which will determine
the quality of the offer comparing with the average
price of the route.
Once default fares are stored, we are at second
phase (see Figure 6), in which it will be possible to
make combinations between these routes. The first
combination process is made with one level of
depth, i.e., we expanded a node in one the level of
depth search. The combinations obtained will be the
result of filtering all possibilities with a variety of
date and flight time restrictions.
At the third phase, we perform a process of
comparing the existing offers in the system and new
combinations of routes processed. This process of
comparison will be made between the price of the
new route obtained, and the average price of the
route of the origin and destination of the new offer
(the new route obtained). If not exist connection
route, would be established a new one directly. On
the contrary, if the route already exists and has a
lower price than average, will be stored and offered
as a possible offer.
After completing this processing of information,
we are at phase four of the architecture. After storing
a new route, the process is repeated continuously.
The search for new routes stops when the price of
the new route is higher than the average price. This
search is based in a deep level determined by the
amount of times of processed routes.
The combination process of possible new routes
is activated when storing new data extracted through
Skyscanner, optimizing with that, the response time
to publish a new offer. The process of publishing
results is present at all stages, listening new
insertions into the database.
After presenting the architecture of our system, we
obtain the knowledge to create the framework where
the publication of flight offers is not limited by the
restrictions that apply airlines to their airfares. This
framework also noted for its ability to offer users the
best flight offers, publishing them in an order that
determines the quality of the offer. Figure 6 is a
diagram of the processes that belong to the system
architecture.
Figure 4: Architecture diagram.
A Web Integration Framework for Cheap Flight Fares
265
In the next section we will give a working example
how the system gives the output to users.
5.2 Problem Definition
Our intention is to implement a system to find the
best offers in existing flights without having to
perform a thorough and manual search based on
routes and dates.
The flight search engine used to perform this
analysis of prices is "Skyscanner", because it does
not include internal rates when searching for flights,
and automatically redirects to the main agents
shopping online (Expedia, Opodo, eDreams, etc.)
and major airline companies.
Doing various searches to check the veracity of
this assumption, we can see that a flight from
Barajas airport (Madrid, Spain) and destination
airport New York John F. Kennedy (NY, USA)
could cost nearly 550€. However, if we do an
exhaustive search performing different route
connections in Europe before crossing the Atlantic
Ocean, we can obtain a price around 400€. In this
case was chosen as the optimal connection Oslo
Gardermoen airport (Oslo, Norway).
Therefore, if we search only the origin and
desired destination, flight search engine will show us
different air routes with a price about around 150€
more expensive.
Another case in which we can see that the search
for lower prices are made according to the fares
restrictions of holding corporations, is the alarming
case to make trips from Ireland, in any of its cities,
and as a destination Puerto Vallarta (Jalisco,
México).
If we search with origin airport Dublin (Dublin,
Ireland) and destination Puerto Vallarta (Jalisco,
México), we get a price almost 900€, performing all
routes with an airline holding, Skyteam, including
flights from airlines like Delta, Aeroméxico and Air
France.
On the contrary, we found another cheapest
option for the same date, but under different
corporate holdings, and also making connections
within Europe before crossing the Atlantic Ocean. In
this case the airport connection between the origin
city and destination would the London Gatwick
airport, where the fare is less than 100€ by Ryanair.
From London Gatwick airport to Puerto Vallarta
with the company Thomson Airways would have a
fare less than 400€. A saving of almost 400 euros!
Therefore, for the same origin and destination,
performing different routes, we can be able to get a
price 50% lower than shown by default search
engine. Such examples can be found in several
destinations, both on flights from Europe to
America, like from Europe to Asia.
In this part of the process of analysis of results, is
where take part the functionality of implementing
optimal search minimum cover algorithms of our
proposal. The intention is to filter and limit searches
to a certain depth, obtaining as a result offers lower
Table 3: Skyscanner search at 06/04/2015.
Origin Destination Departure Arrival Price Company
MAD JFK 29/01/2016 30/01/2016
552€ Delta + Aeroméxico + Air France
JFK MAD 05/02/2016 06/02/2016
Origin Destination Departure Arrival Price Company
MAD OSL 29/01/2016 29/01/2016
173€ B. Airways + Iberia
OSL MAD 06/02/2016 06/02/2016
OSL JFK 29/01/2016 29/01/2016
251€ Norwegian
JFK OSL 05/02/2016 06/02/2016
Table 4: Skyscanner search at 06/04/2015.
Origin Destination Departure Arrival Price Company
DUB PVR 11/04/2015 12/04/2015
855€ United
PVR DUB 25/04/2015 27/04/2015
Origin Destination Departure Arrival Price Company
DUB LGW 11/04/2015 11/04/2015
96€
Aer Lingus
LGW DUB 26/04/2015 26/04/2015
LGW PVR 11/04/2015 11/04/2015
387€
Thomson
Airways
PVR LGW 25/04/2015 26/04/2015
WEBIST 2017 - 13th International Conference on Web Information Systems and Technologies
266
than average price established for a given route, and
after that, making possible combinations of offers
obtained above.
The application of this algorithm would not be
useful for a path when repeating the process, and did
not obtain results in a final path with lower price
than obtained when it made searches by default way.
6 CONCLUSIONS
In our work we conclude that relational databases do
not offer the necessary flexibility to process flight
management problem with optimal solutions, due to
strong structural constraints and a weak dynamic
indexing. We reach the conclusion that Graph
Databases are the best solution for storing and
processing data flight fares and we choose Neo4J.
We review and analyse diverse existing commercial
solutions and tools to implement new features using
data integration between different platforms. To
overcome the restrictions on flight fares was our
main motivation for implementing a data mining
system, in such a way, that we dispose a system free
of restrictions. With this system, we can show to
final customer the best possible flights fares,
combining connections between different airlines
and even different airline holding companies.
Also, the business alliances between airlines hinder
the process to obtain the most competitive offers,
and therefore, when choosing the cheapest way to
travel, these types of alliances would be damaging
the final consumer. It is true that this kind of
enterprise unions provides many facilities to
customers, due to the centralization of the purchase
(Pels, 2001). But, how much would be willing to pay
for this convenience? This is an open question that
we would like to have an answer in a near future.
We present a framework to find the best deals on
airfares, and with them make combinations that have
as results new flight routes, offering to the customer
a trip with a truly affordable price. To the best of our
knowledge this is the first proposal of this new
feature that is missing from all OTA systems.
As future work we propose a new feature
implementated by major GDS, or flight search
companies, to allow a real optimized search.
Resulting that the customers have the option to find
the best flight fare possible for a specific date.
We also must take in consideration the amount of
information processing that this feature needs (IBM
Big Data, 2015), because it would have to perform
an analysis of real-time fares from a massive amount
of data and could easily overload a system.
Therefore, our system should be able to process
flight fares from more than 30M of annual flights,
and the forecast worsens in 2030, can reach up to
60M annual commercial flights.
REFERENCES
AXSES, 2013. AXSES Travel marketing about global
distribution, GDS history and GDS marketing.
Axses.com. Retrieved August 20, 2013, from
http://axses.com/encyc/archive/arcres/arcrates/users2/g
lobalmarketing_about_gds.cf
Binstock, M., 2015. Oracle and the End of Programming
As We Know It. Retrieved April 11, 2015, from
http://www.drdobbs.com/jvm/oracle-and-the-end-of-
programming-as-we/232901227
Guterman, J., 2015. Release 2.0: Issue 11. O’Really Media
http://www.oreilly.com/data/free/release-2-issue11.csp
IBM Big Data. 2015. What is Big Data – United States.
(2015, January 27). Retrieved April 11, 2015, from
http://www.ibm.com/big-data/us/en/
Jonas, D., 2013. Vision 2020: Old Dogs, New Tricks And
The Future Of GDSs. Retrieved November 22, 2013.
Kamel, M. N., 1994. A prototype rule-based front end
system for integrity enforcement in relational
databases: An application to the naval aircraft flight
records database. pp. 713–723.
Pels, E., 2001. A note on airlines alliances. Journal of Air
Transport Management Volume 7, Issue 1, pp. 3–7
Valles, A.J., VanLoy, J.A., 1991. An Expert Auditing
System for Airline Passenger Tickets. In Proceedings
of the The Third Conference on Innovative
Applications of Artificial Intelligence (IAAI-91).
Woodie, 2015. Neo Rides the Graph Database Surge.
Retrieved April 7, 2015.
A Web Integration Framework for Cheap Flight Fares
267