Académique Documents
Professionnel Documents
Culture Documents
METHOD
developed under the Final Year Work or Project of the 5th year of the Electrical
and Computer Engineering graduation course in collaboration with Adaptive IDE
Lda.
July 2005
Supervisor/Orientating Professor:
Professor José Soeiro Ferreira (FEUP)
Project developed by
Filipe Alexandre Camacho ee99191@fe.up.pt
Frederico Vilas Boas ee99136@fe.up.pt
João Bernardo Câmara ee99220@fe.up.pt
INDEX
1.Motivation.......................................................................................6
2.Introduction ....................................................................................6
5.Theory ......................................................................................... 12
7.1 Introduction...........................................................................38
7.2 Description of the search engine..................................................44
7.3 Individual Travel Selection: implementation ...................................45
7.3.1 Construction of the judgement matrixes ....................... 47
8.Testing......................................................................................... 50
9.Conclusions ................................................................................... 52
10.Annex......................................................................................... 55
11.References .................................................................................. 58
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 2
Acronyms and Symbol Listing
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 3
Picture index
Fig.12 – Search engine with selection boxes and hotel listing ..........................45
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 4
Table index
Table 2 and 3 - Example of the two types of matrixes, criteria matrix and
alternatives matrix: pair-wise comparison and local weights ..............22
Table 22 – Conversion between the 0-10 scale and the AHP scale.....................48
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 5
1. Motivation
The origin of this project dates to late February, early March, after
ADAPTIVE, an entrepreneurial company based on Funchal requested an application
for its tourism contents webpage. The idea behind the requested application was a
client-support system for assisting webpage’s visitors in finding the hotel/resort or
other type of tourism lodging that best suited their interests. Such relevant
interests were pointed out like the price tag, facilities available, services and
others. Besides this application should target an individual user that wishes to
select a hotel.
After a first phase of exchanging information between the Final Year Report
students and the company, a proposal for implementing the web application was
made to the Project orientator. Before the acceptance of the proposal, a road-map
was proposed for achieving all the goals listed in the proposal.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 6
2. Introduction
Have you ever wanted to book the best place in your favourite holiday’s
destination that fitted the exact bill for your perfect vacation? That resort or hotel
that has all the services in the perfect spot for assuring the perfect time, all this
while still fitting in your desired price tag?
And what about if you wish to travel in a group that may hold different ideas
concerning their perfect holidays…it’s all about finding the place that can offer the
best solution for all the members. Surely no one expects the impossible, or a
“heaven” sent compromise between all members which may be well out of reach.
But a solution that could approximate all members’ choices is possible and
desirable.
The big problem is that this type of decision is usually disregarded of any
analytic base: of course it’s hard to say to someone who is planning his/hers
holidays to grab a piece of paper and a pencil, and get “analytical” with some kind
of mathematic algorithm.
Rather the challenge is to incorporate this mathematical analysis and create
a proper interface in order for it to be user-friendly and sufficiently accessible to
all kinds of people, computer educated or not.
This is the scenario for which our project was guided, namely a web
application that supported decision making by a user or a group of users. In this
point, research through literature (papers, online documents and major tourism
boards’ web pages) indicated no similar implementations of the same nature,
which gave the green light for further studying of the scenario.
After some studying, it was found that the baseline theory associated with
this project lies on Decision Science, an increasingly important field of Operations
Research, with applications in so many different environments that range from
biotechnologies, medicine, informatics, logistics and management…in fact the
scope of Decision Sciences can be found in almost every sector of global economy.
In our case the implementation will be of course based on information
systems technologies. Such type of system is commonly known as a Decision
Support System (DSS), defined by Sprague and Carlson1 as an ‘interactive
computer-based systems that help decision makers utilize data and models to
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 7
solve unstructured problems’ although many other definitions are available in a
more or less complete way.
Its origins date back to 1965 and treads closely with the evolution of
computers and information systems. Also it is considered that the concept of DSS
became an area of research of its own in the middle of the 1970s, before gaining
intensity during the 1980s.
In the middle and late 1980s, Executive Information Systems (EIS), Group
Decision Support Systems (GDSS), and Organizational Decision Support Systems
(ODSS) evolved from the single user and model oriented DSS.
Beginning in about 1990, data warehousing and On-Line Analytical Processing
(OLAP) began broadening the realm of DSS. As the millennium approached, new
Web-based analytical applications were introduced (Figure 1).
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 8
Fig.2 - The five distinct components of a DSS
This was the main concern in the documentation phase of our project,
finding a mathematical model that effectively introduced decision making for
scenario in hand, firstly considering only an individual and finally for the group
travel selection .
The mathematical model used for this phase of our project was the Analytic
Hierarchical Process (AHP) a popular multiple-criteria decision making tool based
on hierarchical structure. Some of its attracting features, among others, are the
capability of synthesizing qualitative, as well as quantitative info into the decision
making process. More on AHP and its implementation this will be the subject of a
more detailed explanation later on.
The final stage of the project is the implementation phase, the construction
of the DSS using information systems technologies, such as HTML, CSS and PHP4.
These are popular tools in database programming nowadays.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 9
Fig.3 - Operating structure for the DSS: sequence of events
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 10
3. Case Study
In brief…
The goals: build a model that effectively implemented decision support making to
the user. The requirements for the implementation, other than the interface, were
to take the theory model built for an individual and “translate” it into a web
application. An appropriate interface should also be put in place.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 11
4. E-Tourism: Current scenario and trends
The Internet has made possible numerous products that have enhanced the
tourism industry, coming from the typical online brochure of a hotel/tourist
destination to incorporate multimedia and online services in its contents opening
the way for new profit/services possibilities.
In fact, owing to their intangible and digital characteristics, tourism
products may no longer be needed because tourists will communicate directly with
hotels and airlines electronically. Or meaning the tendency to avoid intermediaries
in the process. Despite this only a marginal part of the tourism profit is generated
online.
For e-commerce this means that tourism business activities will have to go
beyond the already present online reservation and offer the customers other value-
added services, if the need of intermediaries is to be reduced. Clearly this will
require innovation and sheer entrepreneurship.
In fact, the slow adaptation to e-commerce innovations leads to high
product similarities and severe price competitions among web site operators, as
well as low total tourism market share, as already stated. Also it has been signified
that a more consumer-oriented web-based tourism information system to support
users in travel-related information search, product bundling, and travel planning,
and so on is strongly desired. This is clearly the next step in e-tourism3.
Another important factor in designing this client customized services passes
also in gaining the trust of the customer and the creation of sufficient incentives
for stimulating the curiosity of customers in the business proposition (Nysveen and
Lexhagen3).
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 12
5. Theory
In this part we will give some insight into the theory behind our project,
concerning all the subjects mentioned in the introduction.
Firstly we have to make additional remarks to our DSS and also to situate it
into the actual scenario of e-commerce and e-tourism. Also we will devote some
space to the theory behind AHP, since it will be one of the cornerstones of this
project.
This will be dealt in the form of exposing the algorithm and then
consequently show the way in which the AHP theory was incorporated into our case
study problem.
4
5.1.1 The scope of DSS in e-commerce
The rapid advancement of Internet and Web technologies and the fast
growth of e-commerce applications in recent years have brought strong impacts on
the strategies and processes of business conductions. Many innovative business
models have emerged in the e-commerce environment such as market-oriented e-
Shop, e-Procurement, e-Auction, e-Mall, Third Party Marketplace, Virtual
Communities, Value Chain Service Provider, Value Chain Integrator, Collaboration
Platforms, Information Brokers, and Trust Service Provider.
Major identified e-commerce characteristics include global markets, virtual
organizations, 24/7 operations quick responses, competitive pricing, secure
transactions, multimedia and hypermedia documents, interactive processes,
personalized and customized services, value-added information, innovative
products and services, etc.
The growing Business-to-Consumer (B2C) applications and increasing market
competition have stimulated the needs for more information-intensive and
decision-oriented online consumer-support systems and services that could
incorporate personalized needs and interests in all searching, deciding, and
purchasing processes. As stated previously, in e-tourism the current trend is for
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 13
tourism information systems to offer extended decision-making support in tourist
travel planning.
It is obvious that the desired e-commerce-oriented consumer decision and
transaction process is relatively more complex than the traditional buying process,
since it may contain online activities such as product search and discovery, product
and vendor evaluation, price and contract negotiation, transaction and payments,
post-purchase services and dispute resolution.
Moreover, when planning and transaction services for consumer groups or
communities are concerned, extended group decision support capabilities should
be developed and provided.
Therefore, how to apply innovative e-commerce related models and
technologies to facilitate the web-based consumer decision and transaction process
that supports individual and group decision making with expert-level qualities
becomes critical for sustaining e-business competitiveness. As a result, more
sophisticated concepts and advanced technologies for designing the consumer-
oriented intelligent decision support system need to be developed to meet the
increasing market demands.
An intelligent consumer-oriented DSS can be generally identified as a web-
based DSS that provides generic and specific application functions, information
resources, model and knowledge computing mechanisms, as well as communication
facilities to efficiently and effectively assist consumers in making personalized and
group decisions through all phases of the decision and transaction process.
Potential business applications of the consumer-oriented DSS range from
online customized shopping, personalized insurance planning, personal financial
and investment portfolio management, to individual or group travel planning.
4
5.1.2 History of DSS implementation in the Tourism industry
Since the very beginning of this project, a thorough search was made via
internet to position our proposed project and confront it with already developed
products in the DSS area.
The search involved mainly online libraries and other scientific database
warehouses, like ScienceDirect.com also the online ACM Portal (Association for
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 14
Computing Machinery). Also the search complemented several tourism portals, like
Opodo (www.opodo.com, a large travel portal for reservations and flights owned by
the main European airliners), Expedia (www.expedia.com, a U.S based tourism
portal with more or less the same scope as Opodo), and Kayak (www.kayak.com, a
U.S online travel search engine). The objective with visiting these previous sites
was to get a feedback on the type of technologies besides the already omni-present
search engine.
Most of these sites indeed already incorporated many up-to-date
technologies, like AJAX, which is an intelligent way to save channel bandwidth by
only refreshing the desired part(s) of a webpage and also CSS, which provides for
some stunning interfaces and easier manipulation of styles inside a web page.
As for examples of web applications tourism focused DSS’s, these were few
and far between. One of the most relevant was A Web-Based Consumer-Oriented
Intelligent Decision Support System for Personalized E-Services 4, which dealt
with presenting an integrated framework for developing web based consumer-
oriented intelligent decision support systems to facilitate all phases of consumer
decision-making process in business-to-consumer e-services applications,
culminating with an example given for e-tourism. This paper, in the literature
review part, indicates that in commercial websites, currently exists ‘some efforts
to assist customers in searching and selecting products and services have been
reported’, or in other words, client-support for decision aiding.
The paper goes on to state some examples, like General Electric Plastics
(www.geplastics.com) which provides datasheets, engineering calculator, and
material selection tools on the company web site to help customers in analyzing
product needs and getting an effective material solution. It concludes this
literature review by stating that ‘although the needs to offer more powerful
capabilities for consumer decision support on the web sites are widely recognized,
the facilities already provided to the consumers are still limited to specific
products and tasks and thus unable to support full-stage and high-quality decisions.
The other part of special interest is the application of the proposed
integrated framework in the e-tourism web application. Several prototype systems
were developed for both the e-tourism and e-investment applications using the
proposed framework and design methodologies, using as background the tourism in
New Zealand. In one of the these there is a system for evaluating package tours in
which consumers select their preferences and weights about destination region,
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 15
trip length, price, accommodation rank, departure date, and features, as well as
specify the level of matching. Users can then submit the request to get a list of
matched tour packages for their inspection.
Another interesting application is a page for designing personalized tour plan
in which consumers can design their own trip plan by selecting and bundling
destinations, hotels, and restaurants in daily basis.
Other page is aimed at group travel: community voting that allows
community members to vote on original and alternative trip plans. Before they
make the vote, users can check the content of each trip plan. After inserting a new
vote by someone, new vote counts of all trip plans appear on the ‘number of votes’
column.
Even another great example of value-added online service: the page has a
tour plan bidding session that allows travel agencies to bid on posted group trip
plans. The time interval for submitting a bid, the current lowest bid, and the name
of the associated bidder are also shown in this page.
Finally another application illustrates a continuing recommended personal
insurance portfolio plan in which insurance types, principal, duration, and premium
are shown in response to a consumer’s need and preferences.
A final note is required to say that these described systems are only
prototype at the moment, after we tried to retrieve more information in the net
for better understanding of features and technologies
Work on the taxonomy of DSS has been conducted since the development of
DSS. As with the definition, there is no all-inclusive taxonomy of DSS either.
Different authors propose different classifications. This classification is in
general done at three levels:
• user level, the level of interaction between user and system,
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 16
• conceptual level refers to the goal for which the DSS was designed,
type of operation, data manipulation…
• technical level, if the system is based in one single computer or
distributed around a large organization
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 17
5.1.4 Taxonomy of the proposed DSS
In this part, we need to clearly specify the aspects or features that will be
used for assessing DSS performance.
One element for further improving/upgrading this project is to put in place a
common framework that will allow for the use of this application (of course with
some differences in its implementation) in other tourism destinations.
This means that the system must have some flexibility in order to adapt to
contrasting destinations, like for example a skiing destination. If we look to the
available literature dealing with DSS’s, one can see that the main drawbacks in its
design are:
• Poor maintainability, that illustrates that a decision-maker
sometimes has to leave the focus on decision making and has to spend
some time and attention in maintenance of the DSS.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 18
• Poor flexibility, which means that DSS’s are often too much
application-specific, with difficulties in updating/upgrading
In fact and given the nature of our proposed DSS, we can consider flexibility
for different tourism destinations as the main factor in evaluating our application.
For the present problem there is a need to address the way in which users
will express their preferences/wishes and evolve from there to a ranked
prioritization of alternatives according to the expressed characteristics.
The Analytic Hierarchical Process (AHP), created by Thomas Saaty13 in the
1970’s, is an excellent tool for optimization procedure in multi-criteria
environment, when several alternatives are presented to the user.
It allows as we already stated the capacity to synthesize both quantitative
and qualitative information into a hierarchical model, by means of pair-wise
comparisons of alternatives of criteria and then of alternatives to the criteria
proposed in the decision problem. The method is comprised of the following steps:
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 19
The simplest AHP model has only three levels, namely Objective, Criteria
and Alternatives. This phase is a fairly important part of the method so as to
represent the decision making problem faithfully. Saaty13 recommends care in
designing the model so that the structure effectively represents the problem in
hands.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 20
Intensity of Definition Explanation
importance
In this step, local weights of the elements are calculated from the judgment
matrices using the eigenvector method (EVM).
The normalized eigenvector corresponding to the principal eigenvalue of the
(judgment) matrix provides the local weights of the corresponding elements.
To do so, there is the need to define the comparison matrixes which makes
the pair-wise comparisons between each criterion of the decision model.
After knowing the preferences of the user by an interface used to capture
the more preferable criterion, these are represented in a matrix which compares
the criteria with respect to the main goal (Fig:4 Table A), calculating the local
weights of each criterion has to the user.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 21
Secondly, the need of comparing each alternative with respect to each
criterion (criterion 1…, criterion n), will give respectively the local weights of each
alternative
However other comparisons must be made in order to capture the relation of
preference between possible alternatives. For that reason other matrixes are
needed in order to calculate the local weights, there are built n+ 1 matrix as n,
number of criterion.
Table 2 and 3 – Example of the two types of matrixes, criteria matrix and alternatives matrix: pair-wise comparison and
local weights
Step 4 – Aggregation of weights across various levels to obtain the final weights
of alternatives.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 22
The final weights are calculated using the local weights of each alternative
regarding each criterion and the local weights of each criterion as shown in the
expression above.
This expression will calculate the priorities which each alternative.
Notice that the sum of all final weights must be equal to a unit (1). This will prove
that the method used is being well followed as well as confirming all the steps
made before.
The final values reached (estimate weights) inform (to contemplate explicit
or implicit knowledge) about the possible alternatives and the way they are used to
satisfy the selected criterion, as well as the importance of these criterion in order
to reach the goal of the better alternative to choose from.
Taking this into regard we have reached a result where we can affirm which
alternative is more preferable from the user point of view.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 23
An arbitrary but generally-accepted as tolerable level of inconsistent
preference scoring with the AHP is less than or equal to 10% of the total number of
judgments.
The process of acquiring the values so we can reach the final result is
divided into 4 steps:
1. Synthesizing the pair-wise comparison matrix
2. Calculating the priority vector for a criterion
3. Calculating the consistency ratio
4. Selecting appropriate value of the random consistency ratio
1st. Calculate the weights with respect to the initial matrix of criterion and its local
weights (criteria weights).
2nd. Calculate the inconsistency index with respect to the criterion weights and the
weights driven in Step 1:
λ max − n
IC =
n −1
4th. Compare the IC ratio with the random index (RI) with respect to the
corresponding n.
IC
RI
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 24
Size of 1 2 3 4 5 6 7 8 9 10
matrix, n
• Verbal to numeric scale conversion – decision agents that use the AHP’s
verbal comparison mode have their judgements automatically converted to
the 1-9 numeric scale, but the correspondence between the two scales is
based on non-tested theory base. For example if A é judged to be weakly
more important than B, the AHP assumes that A is 3 times more important
than B, which may not be the case at all. There is some argumentation that
factor 5 is too strong to denote the notion of strong preference.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 25
and B is 5 times more important than C that should make A 25 times more
important than C; but that is not possible, due to the fact that the scale
goes from 1 to 9.
Even with these drawbacks, the method has great intrinsic value due to the
hierarchical structuring of a problem and the facilitation of the dialogue between
decision makers. In this fashion, the AHP is perfectly adapted as long as its
limitations are taken into account. A good example of this, is the large number of
DSS applications that incorporate AHP as a decision making tool.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 26
Turning our attention to group decision, many refer to AHP ‘as well suited
for this task due to its role as a synthesizing mechanism by means of its capability
to accommodate both tangible and intangible characteristic, individual values and
shared values’ (Lai 8). Also it can help structure a group decision so that the
discussion centres on objectives rather than on alternatives.
PHP
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 27
dynamic Web pages’ as well as to an intuitive syntax which facilitates learning to
anyone with basic programming skills.
Also because it is an open-source product, allows the support of a large
group of open-source developers worldwide. This provides users with excellent
technical support and bugs are found and repaired quickly.
Other feature is the excellent connectivity to most of the common databases
(including Oracle, Sybase, mySQL, ODBC and many others), and its integration with
various external libraries, which allows the developer to do anything from
generating PDF documents to parsing XML and other.
Another key advantage of PHP, when compared to other scripting languages
such as ASP or ColdFusion, is that it is open-source and cross-platform, suitable for
today's heterogeneous network environments.
Given this and according to [9], PHP is accounted ‘as today's fastest-growing
technology for dynamic web pages’ and according to a specialized internet
technology survey (conducted by Netcraft, www.netcraft.com/survey/ ) ‘PHP can
now be found on more that 6 million domains, and is growing at a rate of up to
15% each month’.
SQL 10
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 28
BASIC which, being programming languages, are designed to solve a much broader
set of problems
Compared to general-purpose programming languages, this structure allows
the user/programmer to be less familiar with the technical details of the data and
how they are stored, and relatively more familiar with the information contained in
the data. This blurs the line between user and programmer, appealing to
individuals who fall more into the 'business' or 'research' area and less in the
information technology area.
The original vision for SQL was to allow non-technical users to write their
own database queries. While this has been realized to some extent, the complexity
of querying an advanced database system using SQL can still require a significant
learning curve.
Although SQL is defined by both ANSI and ISO, there are many extensions to
and variations on the version of the language defined by these standards bodies.
Many of these extensions are of a proprietary nature, such as Oracle
Corporation's PL/SQL or Sybase and Microsoft's Transact-SQL.
It is also not uncommon for commercial implementations to omit support for
basic features of the standard, such as the DATE or TIME data types, preferring some
variant of their own. As a result, in contrast to ANSI C or ANSI Fortran, which can
usually be ported from platform to platform without major structural changes, SQL
code can rarely be ported between database systems without major modifications.
There are several reasons for this lack of portability between database systems
such as:
• The complexity and size of the SQL standard means that most
databases do not implement the entire standard.
• Many database vendors have large existing customer bases; where the
SQL standard conflicts with the prior behaviour of the vendor's
database, the vendor may be unwilling to break backward
compatibility.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 29
6. Design of the AHP model
After a thorough description of the AHP method, it’s time to create the AHP
structures that will carry translate the scenarios exposed for the individual & group
travel selection.
We will derive a model for each of the two cases using the already exposed theory.
In interpreting the individual travel selection one should first identify clearly
the elements that we will want to integrate into our structure. From the case study
presented it’s obvious that the alternatives will be hotels or other kind of tourist
lodging. This was made clear when the goals were exposed.
Secondly and since the decision making resides only in an individual, the
goal is to optimize the client’s input preferences for his/hers desired vacation
lodging. The final part is to identify the criteria for which we wish the customer to
do the pair-wise comparison in the 1-9 scale. The issue is identifying the right
parameters that will describe completely our scenario.
This part was carefully accompanied with responsible of ADAPTIVE, through
swapping of views and ideas about the subject: after some discussion, three
criteria items were selected:
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 30
• Price
The price range is an obvious criterion in travel/hotel selection problem.
• Quality of Accommodation
This was the criteria that raised more doubts about its existence. In fact one
can say that price is directly related to the quality of the accommodation and thus
the parameter being useless. This is because the AHP method recommends care on
the selection of criteria based on mutual exclusion of criteria.
However, and after some research around the hotels of Madeira in general,
one can see that for example there were quite a few hotels of 3 stars that had a
very similar price compared to 4 star ones. This was also the case with some 4 and
5 star hotels although in a less significant manner. For this reason and because the
number of stars in a hotel determines certain quality parameters of services and
infrastructures, we decide to include the quality of accommodation in our
structure. The star rating will serve as guarantee of services for the user.
The next step is to define the hierarchical structure, based on the previous:
As we see the structure is a very simple AHP structure with 3 levels. One
final note is about the number of alternatives at our disposal: in the scheme
presented the alternatives are supposedly as many as the user wants.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 31
Let’s not forget that before the optimization procedure there is a search
engine that will limit the alternatives according to the preferences of services,
infrastructures, nearby facilities and/or services etc…the search engine will be
properly detailed in the implementation phase.
The Individual Travel Selection didn’t pose much work on the documentation
due to the fact that it is a classic application of the AHP method, as seen in
numerous books/papers on the subject.
For the Group Travel Selection we wanted to continue to use AHP as it would
permit considerable savings in the implementation of the web application. So the
emphasis was in finding adequate literature that dealt with Group Decision Support
System (GDSS) in a multi-criteria environment preferably in theory for the purpose
of orientating our work in this stage of the project. The first task involved
consulting papers and/or other types of information (web sites, etc…) that
employed some kind of a Group Decision Support System (GDSS).
After this stage several papers were considered, but we decided to base our
implementation of the Group Travel stage on two papers, namely Group Decision
11
Making and Hierarchical Modelling , by Jablonsky and Lauber, Group Decision
Making in Multiple Criteria Environment: A case using the AHP in software
8
selection by Vincent S. Lai, Bo K. Wong and Waiman Cheung and finally
Aggregation of analytic hierarchy process models based on similarities in
12
decision makers' preferences by N.Bolloju.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 32
Fig.6 – The AHP structure adopted in Jablonsky et Lauber
The proposed hierarchy presented here suggests a set of levels for which we
will give a brief description. The article assumes that there is an inherent conflict
in the decision making process, so it proposes that the Level 1 (or goal of the
problem) must be an apparent compromise between the members of group (that
may have different weights in the final decision). The paper later goes on to
quantify this “compromise” into an index.
For our case however the goal of looking for a compromise between all is not
important, so the analytical study presented in the paper on how to reach a
compromise between the members wasn’t considered. So for this matter our goal is
to find the alternative, given each member’s weight in the decision process that
best suits the group’s wishes with no special care for conflict resolution (we
assume that the AHP already considers this implicitly in its implementation for
group decision).
The remaining levels are relatively straightforward: Level 2 represents each
member’s view of the perfect location for their holidays (or Party to the Conflict in
the original structure), Level 3 the criteria (the same as in the individual Travel
Selection) and finally in Level 4 the alternatives available to choose from.
The relations between levels derived from the hierarchy can be converted
into a numerical form and interpreted as follows:
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 33
• Level 2 - evaluation of the importance of the parties to the conflict with
given criterion and the given decision maker) and global priorities
(synthesized from the local preferences) are the direct basis for the final
decision (finding the consensus, ordering of the scenarios, etc.).
In this step we will show the hierarchical structure for the Group Travel
Selection, according to the remarks made in previous step:
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 34
Fig.7 - Proposed Hierarchical Structure for Group Travel phase
As stated before, the AHP algorithm is applied to our problem in very similar
conditions comparing to the individual travel selection: firstly the AHP is applied
individually to each individual.
Secondly, the weight in the decision making process (again for each
member) as well as the aggregation of the individual preferences are introduced
into the problem following the Weighted Arithmetic Mean Method, as shown in
12
Bolloju .
From this document that describes methods for synthesizing a group
decision, two different methods are presented: the geometric mean method (GMM)
and the Weighted Arithmetic Mean Method (WAMM).
The reason why WAMM was chosen is because the several members of the
group may have radical differences in their judgment about the most suitable
alternative. The subject of the aggregation procedure will be detailed below:
In first notice, each one of the group’s members must fill in a text box, with
personal data as well as his/hers preferred location (remember that it can be a
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 35
beach resort or a country cottage for example) for their proposed vacations. Then
each member designates its importance in the group decision (from 0 to 1, meaning
1 the total decision for the group and that the sum of importances cannot be
higher than 1). The data is stored in database tables
Finally a listing of all hotels that comply with the alternatives chosen by all
users is displayed. The users can eventually discuss the alternatives and remove
any of them if they wish so. The remaining alternatives enter the individual AHP.
After this is done, the AHP is implemented in roughly the same way as in the
individual choice. For each comparison matrix the user makes the pair-wise
comparisons between for the three selected criteria. The criteria remain the same:
PRICE, QUALITY and LOCATION.
The only difference is in the matrix that relates the LOCATION criteria with
the alternatives offered. In this case, the matrix follows the type of LOCATION
chosen by this individual alone.
After the insertion of data into the 4 tables per member, prioritization of
alternatives is made for each member. The next step will aggregate the individual
results.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 36
Pairwise comparison matrix
Member 1
Fig.8 – Aggregation of the individual judgments into the group’s judgment matrixes for group decision
with k depicting the kth member of the group, i the line of matrix and j the column
of the aggregate group matrix.
This process is followed for the other matrixes which results in the 4
predicted ones.
After these matrixes are computed, we simply derive the local weights and
the final weights in the same manner as in the individual case.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 37
7. Implementation
7.1 Introduction
In this part we will detail the implementation of our DSS, following the
derived models for the knowledge engine already described in the previous parts.
The 1st stage in the implementation is to build an entity-association model for
describing our database. The E/A model must not though to be considered
definitive since it may not describe all the perceived restrictions, due to its
subjective nature but it will, nonetheless, establish a design approach to our
database.
In the E/A model, the scenario we wish to depict is illustrated by means of
entities and its relations between them. The relations between them are pretty
much straightforward so description of it is relatively scarce.
Also before designing a fully functional search engine we have to consider
the entities/attributes that we will use to describe a hotel or other kind of tourism
lodging. Considering all the possible information about tourism lodging we selected
the following entities (bold) and its respective attributes (underlined):
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 38
views from the hotel are available. Finally there is a classification of
the hotel according to the accesses and location: beach, rural,
mountain and city (called nearby). This classification is not limited to
one single choice: a hotel can be classified as beach and city for
example.
• Types of rooms available: description of rooms and their number
• Food, describes if type of meals (half-board, full board) or the
customer has facilities for self catering.
• Facilities, describes the facilities available in the hotel (for example,
tennis courts, swimming pools etc…)
Entities
Chain
Fields Type Null? Links for Commentaries
namechain varchar(30) No Name of the hotel chain
Facility
Food
Field Type Null? Links for Commentaries
namefood varchar(20) No type of food
Location
Field Type Null? Links for Commentaries
Type of access that an
access varchar(15) Yes
hotel or hotels gots
Type of view that an
views varchar(10) Yes
hotel or hotels gots
If it nearby or close
nearby varchar(15) Yes
from a place
The address of a
street varchar(50) Yes
certain hotel
Postal code. The first
zip1 varchar(4) Yes
four digits.
Postal code. The last
zip2 char(3) Yes
three digits.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 39
Municipally –>
municipally varchar(20) Yes Municipality
name
Municipally
Field Type Null? Links for Commentaries
name varchar(30) No Name of the municipality
Roomtype
Links
Field Type Null? Commentaries
for
Name of the type of
nametype varchar(25) No
room
Service
Field Type Null? Links for Commentaries
nameserv varchar(20) No Name of a service
Typeserv
Links
Field Type Null? Commentaries
for
If it is paid of free the
name varchar(10) No
service
Unit
Field Type Null? Links for Commentaries
nameunit varchar(30) No Name of the unit
numstar int(11) Yes Number of stars
type varchar(10) No Type of unit
phone1 varchar(21) No
phone2 varchar(21) Yes
fax varchar(21) Yes
website varchar(60) Yes
email varchar(40) Yes
Minimum Price that
pricemin int(30) Yes
hotel can offer
Maximum Price that
pricemax int(30) Yes
hotel can offer
chain -> Name of the chain that
namechainu varchar(30) No
namechain a hotel belongs
location ->
streetunit varchar(30) No The address of the unit
street
Ranking 0 to 10 for the
sea int(11) No implementation of AHP
concerning Location
Ranking 0 to 10 for the
mountain int(11) No
implementation of AHP
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 40
concerning Location
Ranking 0 to 10 for the
center int(11) No implementation of AHP
concerning Location
User
Links
Field Type Null? Commentaries
for
The user will have a certain
ticket that will be used for
ticket int(6) No identifying all users at will be
grouped on Group Decision
Making
Associations
Choose
This table was created to keep the hotels that were chosen on Hotel Search and all
the ranking values that were used for the AHP method.
Chooses
A ticket is going to be attributed for every user that will associate with the
unit he has chosen.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 41
ticket int(6) Não user -> ticket
got
Field Type Null? Links for Commentaries
Nameunitgot varchar(16) No unit -> nameunit
service ->
Nameservgot varchar(20) No
nameserv
typeserv varchar(5) Yes
gotan
Field Type Null Links for Commentaries
nameunitgotan varchar(16) No unit -> nameunit
roomtype ->
nametypegotan varchar(25) No
nametype
gots
Field Type Null? Links for Commentaries
nameunitgots varchar(16) No unit -> nameunit
namefoodgots varchar(15) No food -> namefood
pricefood decimal(5,2) Yes
offers
Field Type Null? Links for Commentaries
nameunitoffers varchar(16) No unit -> nameunit
facility ->
namefacil varchar(20) No
namefacil
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 42
Fig.9 – Database’s E/R model
The database technologies used were SQL, PHP 4.0. One idea already
present for future upgrading this DSS is to use new technologies like AJAX and
SQLite which will provide better usage of resources as well and better
interface/usability from the user’s point of view. Other powerful tool for upgrading
the interface is RICO, an open source JavaScript library for creating rich internet
applications, compatible with AJAX.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 43
7.2 Description of the search engine
The user when accessing the web application will encounter an interface
which will be divided on two important sections:
1. Selection of attributes
2. Display of hotel units (according to selected attributes).
In the top-left one can find the price scroll bar: the user will use it to define
the upper limit for the price tag in the hotel search.
The user can also choose hotels of a particular star rating: this selection is
made on the following text box right below the price scroll bar. One might wonder
if this will affect the AHP method since one of the criteria is the Quality. In fact if
a selection of hotels that has the same star rating enters the attribute comparison
all the derived local weights in the pair-wise comparison for Quality for each hotel
are the same, not influencing in pervious manner the derived global priorities (for
the Quality criteria) that will then follow to the final solutions.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 44
The other selection boxes follow from our description of the database
entities and attributes indicated earlier such as the access, municipality of the
desired hotel, desired facilities and services, views from the hotel or type of food
services offered by the hotel.
The initial step is to illustrate all the hotels from the database; the next step
is when an attribute is chosen by the user a new request will be asked to the
database by filtering a new search.
After the display of the hotels, the user has an option of eliminating any of
listed hotels on the page. In any case the user must select the alternatives that he
wishes to go ahead using the AHP.
In the next step, the AHP model is introduced. We will describe separately
the implementation for the two mentioned case (individual and group travel).
In the individual case, the search engine used will be equal to one described
previously. So in this part we will detail the implementation of AHP method to our
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 45
case, following the established design. The 1st part is to introduce the pair-wise
comparison between the criteria.
Some notes: the scale is not in the usual form of the 1-9 scale as described
earlier but varies between -9 and 9. This is because we need to differentiate the
relation between the pair-wise: for example, if we say that A is -5 compared to B
according to our conversion means B is 5 compared to A, in the 1-9 scale.
After this step, the information needed to accomplish the method is
complete. The next step is to perform the pair-wise comparisons between
alternatives regarding each criterion proposed. All data is contained in the tables
of the database, and classified in a 0-10 absolute scale.
For achieving the 1-9 AHP scale (for the pair-wise comparison of
alternatives) we will use the numeric difference between the values of each
respective attribute regarding LOCATION, PRICE and QUALITY for each of two
alternatives. This difference will then be mapped into the 1-9 scale by means of an
appropriate table of conversion that will be described in next step. After this, the
matrixes are filled in with the appropriate values taken from the lists saved on the
database.
As already mentioned, we will have in any case 4 matrixes, one comparing
criteria vs. goals the other three are the judgements matrixes (pair-wise
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 46
comparison of the alternatives with respect to the criteria). These matrixes will all
be square (number of rows equal to number of columns).
As stated earlier the user input ends at the pair-wise comparison of the
criteria. The values that we are going to input into the following matrixes
(judgement matrixes) are retrieved from the hotel lists in the database.
One problem was how to interpret the various prices of the different types
of rooms available. To make matters worst the type of rooms differed quite
significantly which made comparison of these difficult and also most importantly
the prices varied quite a lot.
After some discussing we decided that the minimum price of each hotel
should be the one to be used. This decision was made after we observed that the
minimum price offered a good indication of the hotel’s general price tag.
The next point is to explain the conversion from the 0-10 scale to the AHP’s
1-9 scale. The problem arises from the fact that the 1-9 scale is a relative scale,
i.e. a classification on the 1-9 scale only has meaning if we compare two
alternatives at a time. Because of the nature of this scale, for each time that the
AHP method is applied to the selected alternatives, we classified the prices in 0-10
scale in the following manner:
The hotel with the most expensive price is classified with 10. All the other
hotels with remaining prices are classified using a simple linear transformation.
This procedure is followed every time that the user enters the chosen alternatives
into the attribute comparison.
The following issue is: two alternatives with values in the 0-10 scale that we
need to convert to the 1-9 scale. Since we are dealing with an absolute scale (the
0-10 scale) we can say that for example that if A has 10 as its price value and B
only 5, we can say that A is twice more expensive or that the difference between
them is 5. This difference will be used to convert into to the AHP 1-9 scale, by
using the following table:
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 47
Numeric Difference AHP Description
between alternatives conversion
(in 0 -10 scale) (1-9 scale)
4
5 If the difference is considered to be relevant (4,5 or 6)
5
we say that one is more important than the other
6
The difference between alternatives are quite
7 noticeable, making one much important than the other
7
9
9 The difference between alternatives are so high, that
10 one says that one alternative is extremely important
compared to the other
Table 22 – Conversion between the 0-10 scale and the AHP scale.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 48
In this case the minimum value of the 0-10 scale (0) means that the hotel
has direct access to a specific location while, for example, 10 would mean that the
hotel is very far from that location.
Example:
Pestana Carlton Hotel is a 5 star hotel with the following location attributes:
1. Sea – 0
2. City Centre – 2
3. Mountain – 8
This classification means that the hotel has direct access to sea, is close to the city
centre and is far from the mountains.
The 0-10 scale was adopted for conferring consistence and equal conversion
for each of the three criteria. Of course this was also the case with the Quality
criteria.
For each time that the alternatives are entered into the Attribute
Comparison the application assigns 10 to the one with the highest star rating. All
the remaining star rating values are mapped into the 0-10 scale.
As it was the case with the previous criteria, the conversion table is used to
get to the values in the AHP scale needed for filling the QUALITY matrix.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 49
8. Testing
In the first step, we had some errors in the insertion of the values which was
quickly solutioned. Also the application effectively converted the values sent by
the database in the 0-10 scale into the AHP’s scale.
For the second step, as mentioned, we used the Expert Choice for assessing
the accuracy of our results. The software requires building a structure and making
pair-wise comparisons between criteria and between alternatives regarding a
criterion.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 50
This was done using the data inserted into the application’s tables. For the
example depicted 3 hotels were chosen:
• Pestana Carlton Hotel, a 5 star hotel, with direct access to sea, quite
close to the centre but far from the mountain area. The minimum
price is 135 €.
• D.Pedro Garajau, a 3 star hotel, quite close to the sea, away from city
centre and far from the mountain area. The minimum price is 33 €.
• Hotel da Ajuda, a 4 star hotel, close to the sea, not distant to the
centre but far from the mountains. Minimum price is 72 €.
Fig.14 - Testing our implementation using the Expert Choice: ranking of alternatives
The results achieved are exactly the same as the ones given by our
application which means that the AHP method was correctly followed in the
implementation process. The application example follows in electronic format
inside the CD, with all details, about the pair-wise comparison made for criteria
and alternatives as well as for the consistency analysis.
Finally we need to check the consistency values given by our
implementation. The software also performs the inconsistency ratio for each of the
pair-wise comparison matrixes.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 51
The Expert Choice gave very similar results to the ones obtained in our
application. The differences were minimal and are probably related to the
rounding of intermediate values.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 52
9. Conclusions
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 53
more valued than Price. There’s obvious inconsistency in the process of this
judgment*.
In conclusion, these are strengths of the method which prompted its
selection as the decision-making tool of this project.
Also the testing done on the previous chapter validated our approach of
implementing the AHP in a model-based framework for our database: the results
obtained from our application compared with those of the Expert Choice software,
yielded the same values. The same applies to the consistency analysis, as stated in
point 8 of this report.
Future upgrades to this project relate to the use of more powerful database
related technologies, such as AJAX, which is a smart way to reduce bandwidth by
refreshing only the sections of the webpage selected by the user as opposed to the
refreshing of all the webpage.
Also the interface would be greatly improved by integrating JavaScript
libraries for creating enriched applications inside the webpage. The use of SQlite
would also allow considerable gains in interface/usability aspects.
As for the application’s flexibility for different holiday destinations other
than Madeira’s tourism, we can say that is inevitable that some work would have to
be done in order to adapt it to, for example, a skiing destination.
The main advantage is that the model framework (application of the AHP
method) would be left almost intact. However redesign would have to be done on
the interface and on the E/R model of the database (defining different entities,
attributes and associations if needed).
* Note: testing these inputs results in an inconsistency index well over 3 for the example where a
consistent decision should result in a CR < 0.1.
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 54
10. Annex
Criteria
• Land price
• Distance to suppliers
• Technician’s Quality
• Labour Cost
Solution:
Hierarchical Structure
Best Location
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 55
Normalized Matrix and respective local weights
Criteria LandValue Distance Tech.Qual. Labour Cost Local Weights
Land Value 0.0882 0.0852 0.0732 0.2143 0.1153
Distance 0.7059 0.6813 0.7317 0.5000 0.6547
Tech.Qual. 0.1764 0.1362 0.1463 0.2143 0.1683
Labour Cost 0.0294 0.0973 0.0488 0.0714 0.0617
Sum of input
values from 11.3333 1.4679 6.8333 14.0000 1.000
the previous
table
Ranking of alternatives:
L1 = (0.1153) × 0.5222 + (0.6547) × 0.2364 + (0.1683) × 0.1524 + (0.0617) × 0.2014 = 0.2531
L 2 = (0.1153) × 0.0955 + (0.6547) × 0.0623 + (0.1683) × 0.7208 + (0.0617) × 0.6806 = 0.2151
L3 = (0.1153) × 0.3823 + (0.6547) × 0.7013 + (0.1683) × 0.1268 + (0.0617) × 0.1180 = 0.5318
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 56
Consistency check
Step 1: Calculate the weights with respect to the initial matrix of criterion and its
local weights (criteria weights)
1 1/ 8 1/ 2 3 0.1153 0.4664
8 1 5 7 0.6547 2.8505
=
2 1/ 5 1 3 0.1683 0.7149
1 / 3 1 / 7 1 / 3 1 0.0617 0.2498
Step 2: Calculate the inconsistency index with respect to the criterion weights and
the weights driven in Step 1
1 n weights _ driven _ from _ step1
λmax = ∑
n i =1 criteria _ weights
λmax − n 4,1738 − 4
IC = = = 0.0579
n −1 3
Step 4: Compare the IC ratio with the random index (RI) with respect to the
corresponding n
IC 0.0579
= = 0.0644 ( ≤ 0.1 ), which means that the decision is consistent.
RI 0.90
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 57
11.References
1
Sprague, R. H. and E. D. Carlson, Building effective decision support systems,
Englewood Cliffs, N.J., Prentice-Hall, 1982
2
Marakas, G. M., Decision support systems in the twenty-first century, Upper
Saddle, River N.J., Prentice Hall, 1999.
3
Nysveen H. & Lexhagen M. (2001), Swedish and Norwegian Tourism Websites: The
Importance of Reservation Services and Value-added Services, Scandinavian Journal
of Hospitality and Tourism, vol. 1, nr 1, pp. 38-53, 2001
4
Yu, Chien-Chih, A web-based consumer-oriented intelligent decision support
system for personalized e-services, Proceedings of the 6th international conference
on Electronic commerce, October 2004
5
Gachet, A. (2001), A Framework for Developing Distributed Cooperative Decision
Support Systems - Inception Phase, in Boyd E., Cohen E., and A. Zaliwski (editors)
Proceedings of the 4th Informing Science Conference, June 19-22 Krakow, Poland
6
Machado, Eduardo, Monteiro Gomes, Luis Evaluation of strategies in Services
Marketing: A Multicriteria Approach, Revista de Administração Mackenzie Brazil,
Ano 4, n.2, p. 61-85, Universidade Presbiteriana Mackenzie, 2005
7
Goodwin, P., Wright, G. Decision analysis for management judgment. 2. ed. Nova
York, Wiley, 2000
8
Lai, Vincent S., Wong, Bo K., Cheung, Waiman, Group Decision Making in Multiple
Criteria Environment: A case using the AHP in software selection, European
Journal of Operational Research, Volume 137, Issue 1, 16 February 2002, Pages
134-144
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 58
9
PHP Scripting Language,(Online) available at http://www.zend.com/zend/
aboutphp.php, July 2005
Note: Zend Corporation was founded by the creators of PHP
10
SQL – Wikipedia, The Free Encyclopedia (Online) available at
http://en.wikipedia.org/wiki/SQL, July 2005
11
Jablonsky J., Lauber J., Group Decision Making and Hierarchical Modelling,
Proceedings of the 12th MCDM Conference, Hagen, Germany, Spring 1997
12
Bolloju, N. Aggregation of analytic hierarchy process models based on
similarities in decision makers' preferences, European Journal of Operational
Research, vol. 128, no 3, 499-508, 2001.
13
Saaty, T.L.,The Analytic Hierarchy Process, RWS Publications, Pittsburgh 1990
FEUP/LEEC-PSTFC – A Decision Support System for Hotel Selection using the AHP method - July 2005 59