IMPROVING THE SEARCH AND CATALOGUING OF ITEMS IN
C2C E-COMMERCE PORTALS
Antonio Gallardo, Jose Jesus Castro-Schez, Milagros Hazas
Escuela Superior de Inform
´
atica, Universidad de Castilla-La Mancha, 13071, Ciudad Real, Spain
Juan Moreno-Garcia
Escuela Universitaria de Ingenie
´
ıa T
´
ecnica Industrial, Universidad de Castilla-La Mancha, 45004, Toledo, Spain
Keywords:
Fuzzy cataloguing, fuzzy search and e-commerce.
Abstract:
The business achievement among consumer via e-commerce is getting more important at the present time. In
this paper, we propose to make use of fuzzy logic with the aim to improve the search and cataloguing of goods
and services in Consumer-to-Consumer electronic commerce (E-commerce) portals (e.g. ebay). These portals
are the media through most the electronic transactions among consumers are conduced today. We suggest a
method that tries to adapt to users’ real needs. It allows to buyer carry out searches in an imprecise way and
to the seller to deal with catalogues of items (goods or services) described also in a lacking exactness way.
1 INTRODUCTION
In the last decade, the coming into sight and consol-
idation of the World Wide Web, that involves the es-
tablishment of a competitive environment, has forced
to the firms to develop their sites to a high level of
sophistication and integration (Kowtha and Choon,
2001; Laudon and Laudon, 2005) even allowing the
achievement of business all around the world (Tur-
ban et al., 2000). This fact has caused the appearance
of a new infrastructure to this new business paradigm
known as electronic commerce (e-commerce). It can
be defined as any business that is transacted electron-
ically (Cameron, 1997). The e-commerce processes
and technologies have introduced new ways of doing
business (Chaudhury and Kuilboer, 2001).
Consumer-to-consumer (C2C) e-commerce re-
lates to any business where the transaction occurs be-
tween two consumer who negotiate trying to reach an
agreement or compromise. In spite of the majority of
the commercial transactions are made in traditional
ways yet, many consumers and firms which use In-
ternet to perform commercial activities are appearing
and they are obtaining benefits they could not obtain
using traditional ways. Nowadays, C2C e-commerce
portals are reaching a big summit but the full devel-
opment of these portals and their use needs the de-
velopment and/or integration of methods which fa-
cilitate the search and inclusion of items (goods or
services) in catalogues. In this way, the accomplish-
ment of commercial activities between consumer will
be strengthen.
The great number of portals dedicated to C2C
e-commerce use lexicographic objects arrangements
and searches (i.e. mp3 player, laptop) or direct
searches of the desired commercial model (i.e. iPod
nano, Vaio VGN-SZ2HP/B). Once, a set of items have
been found in the catalogue, they usually allow to give
a order to this set by means of the specification of the
values of some characteristics (e.g. price).
These portals do not allow to us the search giving
only a description of the item by means of the spec-
ification of a set of characteristics, theirs values, the
importance or relevance of these characteristics in our
search and even less do it in an imprecise way.
The lexical search has the problem that if the seller
does not mention in the item textual description an
important aspect when he includes it into the cata-
logue, it will be more complex, difficult and possibly
unsuccessfully the search. Other characteristic which
defines these kinds of searches and arrangements is
the use of synonymous of the entry given by the user.
Then the search engine could find and show items
whose category is totally different with the category
of the item that user is searching. This problem ap-
pears because in the textual description of these items
13
Gallardo A., Jesus Castro-Schez J., Hazas M. and Moreno-Garcia J. (2007).
IMPROVING THE SEARCH AND CATALOGUING OF ITEMS IN C2C E-COMMERCE PORTALS.
In Proceedings of the Ninth International Conference on Enterprise Information Systems - SAIC, pages 13-19
DOI: 10.5220/0002353300130019
Copyright
c
 SciTePress

have been used synonymous. This could cause that
user do not find quickly what he really wants and he
leaves the search. In this kind of portal stand up ebay
(Ebay, ), which is the most used, or froogle (Froogle,
), which has a big search engine.
Others portals, as the Spanish compras (Compras,
), add improvements such as the items arrangement
by mean of a common and important characteristic
and the search by mean of a filter which is usually the
price.
The search based on commercial models force to
the user to have a high knowledge of the market of
the item category. He must enter the model that he
desires. This is the most precise method of search,
but only a little group of users could use this portal
successfully. This fact decrease the number of possi-
ble buyers and makes the portal less attractive, so the
amount of sales the portal could obtain will decrease.
In this paper, we propose to make use of the fuzzy
logic to allow to the user describe items giving theirs
characteristics and values in an imprecise way and we
suggest a method to make searches more precise. It
has been designed to satisfy the users’ needs with the
aim to carry out searches more extensive and detailed.
The remainder of the paper is organized as follow.
Section 2 shows how is the interaction with the portal
and the users’ requirements with regard to the search
and cataloguing of items. In Section 3 how to cata-
logue an item and how the search process is carried
out will be detailed. An example will be presented
in Section 4. Finally, the conclusions obtained dur-
ing the design and development of this portal will be
presented.
2 PORTAL REQUIREMENTS
The portal suggested in this paper will be used in C2C
e-commerce. It will allow to catalogue and search
items through the specification of their characteristics
and theirs values given in a imprecise way. Also, we
allow to associate information about the relevance of
each characteristic according with user’s preferences.
In this way, we try to obtain a list of items arranged
that will be given to the user as a recommendation.
This allow to the buyers to find easier what they are
searching.
To establish the users’ requirements, first we
should determine what users are going to interact with
the portal and how they will do. In this sense it can be
distinguish three different users:
• Administrator: He manages and defines the dif-
ferent items categories and the characteristics of
each category. Administrator should check that all
categories are well determined and that the char-
acteristics which define to each category will be
useful to distinguish the items belonging to that
category. An important part of the portal success
depends on it.
• Seller: He enters items into the catalogue which
could be obtain in exchange for payment by buy-
ers. Seller will describe each item by means of a
characteristics set which will be used to catalogue
it. Seller must put so much emphasis as possible
defining the object since it will be easier find a
buyer. Moreover, the sellers choose the way the
item will be offered, that is by means of auctions
or directly sold to the first buyer.
• Buyer: He enters a specification of desired item
and makes a search with the aim to get a list of
items recommended. A specification consist of a
set of characteristic, the values of that characteris-
tic desired and the importance or relevance asso-
ciated to each characteristic.
Due to the different users that interact with the
portal and the different way to do it, any suggested
portal design should consider the following points
with regard to the use of characteristic and values:
• The administrator may define a new class of
items. To do it, he should determine a new class,
the characteristics that define to this class and the
range of values that they could take.
• The administrator may modify one class previ-
ously defined, adding or removing characteristic
and re-defining the range of values that these char-
acteristics could take.
• The sellers may define the item they insert into the
catalogue easily using non precise values. In ad-
dition to this they should take part in the definition
of items classes and the characteristics and values
that define them.
• The buyers may give specifications in which the
values the characteristics take are given in a natu-
ral way for them, usually in a non precise way.
• The buyers may associate weights to the charac-
teristics in order to increase or decrease its impor-
tance or relevance when a searching is carried out.
• The buyers may use values ranges of a character-
istics when they make a search. The use of ranges
increases the amount of objects closed with the
object desired.
The suggested portal will include others important
features of C2C e-commerce portals (ebay) such as:
• It allows to make lexical searches and by commer-
cial models.
ICEIS 2007 - International Conference on Enterprise Information Systems
14

• It incorporates a message service to allow the
communication between users. For example, it
will be used when a user need to clarify different
aspects of a particular item.
• It includes a register and manage users system
which associates each nickname to a real user.
This information will be used to allow to users get
in touch or communication with.
• It allows to control items sale or auction time.
• It has a system of users votes with a double aim,
on the one hand to associate to each user a respect
of the community rate, on the other to know the
prestige of each one of the users inside the com-
munity that uses the portal.
• It registers the sales of the portal with the aim to
allow to the administrator use this information to
improve the detail of some categories items.
We must point out here, that the specifications and
descriptions are acquired directly from the users. In
the next section, it will be explained how to allow this
freedom when they communicates the characteristics
and values.
3 CATALOGUING AND
SEARCHING ITEMS
The main feature of the suggested method is that it
gives license to the sellers when define an item and to
the buyer when specify the characteristics of the item
which they want to search. Therefore, they could use
the following data types, which we have considered
to be the most common (Castro-Schez et al., 2004b):
• Continuous or numerical, they can take values in
the scale of real numbers (e.g. the house surface
in m
2
or the price of a house).
• Graduated or ranking, they only can take values
from a finite set of values (e.g. the number of
rooms of a house or the house condition).
• Ordered-discrete or ordinal, they can take literal
strings as value. However, these values can be
arranged according to some judgement (e.g. the
district where is placed a house, it can be arrange
having into account where we prefer live).
• Unordered or nominal, they can take any sequence
of words as value and it is not possible arrange
them because there are infinite possibilities (e.g.
the street in which is located a house).
• Boolean or logic, they can take only two values,
i.e. Yes or Not. (e.g. if a house has connection to
Internet or not).
We need a representation which allows all the pre-
viously mentioned values to be described. This repre-
sentation must have the following properties:
• Mixed, it must allow to represent vague and pre-
cise values. The human mind presents this dual-
ity. The uncertainty could appear when a numeri-
cal characteristic is valued in an approximate way
or several values are used to value a characteristic
because its precise value is not known.
• Flexible, it must can hold any value used by the
user (administrator, seller or buyer).
• Computable, it must can be manipulated by a
computer program.
We will make use of the trapezoidal function used
in the fuzzy logic field. This function has four param-
eters (a, b, c, d) and it allow us to represent the con-
crete and fuzzy values. The interval [b, c] establishes
the set of values that belong to the fuzzy value in a
grade equal to 1. The values [a, b) and (c, d] establish
the values that belong to the fuzzy value in a grade
less than 1.
The parameters used to represent each of the types
of characteristics before mentioned can be consulted
in (Castro-Schez et al., 2004a).
Next, we must establish the method to insert and
search items in the portal. Before explain how is car-
ried out the insertion and search of a item into the
catalogue, we need to define some previous concepts.
3.1 Previous Concepts
A catalogue, noted as
C , is defined as a set of classes
of items, C
i
.
C ={C
1
, C
2
, C
3
,...,C
n
}
A catalogue is a not static element. The classes
that define it are not definitive and can be added, mod-
ified or removed from it.
A class consist of a set of items e
j
, that is:
C
i
={e
1
, e
2
, e
3
,...,e
k
}
Each class C
i
is defined by a set of variables or
characteristics V
C
i
(from now we will refer to them as
variables) so that:
V
C
i
={v
1
C
i
, v
2
C
i
, v
3
C
i
,...,v
m
C
i
}
The variables v
j
C
i
that define a class C
i
must be
representative of it, being of use in differentiating the
items e
j
that belong to the same class (e
j
∈ C
i
).
Each variable v
j
C
i
has a definition domain where it
can take values, we note it as DDV
j
C
i
. This set, DDV
j
C
i
,
determines the values that can take an object e
j
of the
class C
i
according to the variable v
j
C
i
.
IMPROVING THE SEARCH AND CATALOGUING OF ITEMS IN C2C E-COMMERCE PORTALS
15

The definition set of a variable v
j
C
i
, DDV
j
C
i
is de-
fined as:
DDV
j
C
i
={A
j1
, A
j2
, A
j3
,...,A
jk
}
with A
ji
being a fuzzy set defined by means of a trape-
zoidal function with parameters a, b, c and d. These
parameters are established depending on the type of
value that they represent.
The definition set of a class, noted as DDV
C
i
, is
defined as:
DDV
c
i
={DDV
1
c
i
, DDV
2
c
i
, DDV
3
C
i
,...,DDV
n
c
i
}
We can associated with the catalogue
C the cata-
logue definition domain set, we noted it as DDV
C
, and
we define it as the union of the definition domain sets
DDV
C
i
of each C
i
∈
C , that is:
DDV
C
=
n
i=1
DDV
C
i
In this paper, it is not important the way of ob-
taining these sets,
C and DDV
C
, though we show
how could be acquired each C
i
and DDV
C
i
. We sug-
gest make use of a semi-automatic method to acquire
the variables V
C
i
, together with the values the vari-
ables can take DDV
C
i
from an expert human present in
(Castro-Schez et al., 2004a). We may use this method
for each C
i
∈
C and obtain DDV
C
.
3.2 Assessing Similarity
Initially, the seller or the buyer must give a detailed
description of the item to be inserted into the cata-
logue or searched to be bought, respectively.
The description of such item, noted as o
i
, con-
sist of a set of features that it has or should have
(V = {v
1
, v
2
, . . . , v
k
}) and the set of values C
v
i
each
feature v
i
could take (C
v
= {C
v
1
, C
v
2
, . . . , C
v
k
}).
Moreover, the variables that are used to describe
the item o
i
could have not the same importance for the
seller or buyer who describes it. They has to indicate
the importance of each variable v
i
in V (∀v
i
∈ V).
Therefore, the description of the item o
i
must as-
sociate to each variables v
i
a importance P
v
i
, that es-
tablish the importance that it has for the user that vari-
able. Each P
v
i
takes his value from the interval [0, 1]).
The greater the value P
v
i
the greater the importance of
the variable v
i
. This allow to pay major attention to
those variables more important for the user when the
search is carried out (Castro-Schez et al., 2005).
To give more freedom to the user at the moment of
making searches, the user can include in the descrip-
tions of the item ranges of values. This option is not
available for all the types of variables.
The use of ranges is allowed in continuous, grad-
uated and ordinal variables. To give a range in contin-
uous and graduated variables the user must insert two
values, those that define the limits of the range. To
use a range in ordinal variables implies to enumerate
of set of literal that the user wants in his ideal object
(e.g. the city districts set where the user would live).
The item o
i
description quality will allow to car-
ried out searches more or less precise and find items
similar to o
i
into the catalogue. In this way, the search
can be adjusted in the catalogue to the real needs of
the user.
We must search the items into the catalogue
C that
will be more similar to the item o
i
. It implies to search
into each class C
i
of the catalogue
C (∀C
i
∈ C ).
The definition domain set of a class C
i
, i.e. DDV
C
i
,
establishes a space of reference in which are placed
all the items e
j
that belong to the class (e
j
∈ C
i
) (see
Fig. 1).












•


•


•


•


•


•


•


•


•


•


•


•



ο



Figure 1: C
i
space of reference.
To search into a class, C
i
, implies to access to the
space in which are defined the items belonging to this
class (DDV
C
i
) and study the proximity of o
i
with the
items e
j
that are placed in this space (e
j
∈ C
i
). The
items e
j
closely located to the item o
i
will be the can-
didates for being bought. On the other hand, if we are
cataloguing the item o
i
, we will make it in the class C
i
whose items are closest to it.
As it can be observed, the search implies to study
the distance between each item of the space, i.e. e
j
,
and the item o
i
depending on the set of variables used
for defining both items. Therefore, it is necessary to
use a measure that provides the above mentioned in-
formation to us and that could be used in a uniform
way without depend on the variable type.
ICEIS 2007 - International Conference on Enterprise Information Systems
16

Since all values that a item takes depending on a
variable v
i
are represented as trapezoidal numbers, the
local distance between two items e
j
and o
i
accord-
ing to a variable v
i
could be assessed by means of
a measurement based on the calculation of the area
existing between the two fuzzy values, A and B that
take these items, e
j
and o
i
, respectively, in the vari-
able v
i
. Such measurement and its properties was sug-
gested in (Castro-Schez et al., 2004a) and we refer to
it as d(e
j
, o
i
, v
i
). In this paper, we use the normalized
measurement of distance d
N
(e
j
, o
i
, v
i
) (Castro-Schez
et al., 2004a), however, any measure d that establishes
the proximity or separation between fuzzy sets rep-
resented by means of the above mentioned functions
could be used.
The global distance between two items e
j
and o
i
is
calculated as sum of the distances local between those
items according to the set of variables used to define
both items. That is, if e
j
∈ C
i
and o
i
is described by
means of set of variables that we call V, then we must
have into account the set of variables V ∩V
C
i
. In this
point, we suppose that both sets have the same termi-
nology. If this was not the case, a method must be
study to obtain it.
The global distance between two objects e
j
(with
e
j
∈ C
i
) and o
i
, noted as D(e
j
, o
i
), will be calculated
as:
D(e
j
, o
i
) =
∑
∀v
i
∈V∩V
C
i
d
N
(e
j
, o
i
, v
i
) × P
v
i
(1)
The importance associated to each variable P
v
i
is
considered in this calculation.
Once the distance between o
i
with each item
e
j
∈ C
i
, D(e
j
, o
i
), has been calculated for all classes
of the catalogue (∀C
i
∈
C ), the system returns a list
arranged by the distance value with a set of items clos-
est to o
i
. The recommended item will be the one with
the lowest distance value.
The searching process in pseudocode is the fol-
lowing:
•
Input: C , o
i
( V = {v
1
, v
2
, . . . , v
k
} ,
C
v
i
= {C
v
1
, C
v
2
, . . . , C
v
k
} and
P
V
= {P
v
1
, P
v
2
, . . . , P
v
k
} ).
• Output: L = {e
i
, . . . , e
j
} | e
i
∈ C
i
and C
i
∈ C .
1. For each
C
i
∈ C do
(a) For each
e
j
∈ C
i
,
i. Calculate
D(e
j
, o
i
) ,
2. Arrange
e
j
according to the value D(e
j
, o
i
)
in L .
3. Recommend the object
e
j
with the lowest
value
D(e
j
, o
i
) .
To make more efficient the search we could de-
termine a subset of classes
D (D ( C ) in which to
carry out the search. The set
D will consist of classes
C
i
belonging to
C that possess a number of variables
commons to the specification o
i
great than mean (τ).
That is,
|V
C
i
∩V| ≥ τ
Moreover, the user could also select the class into
the which he want to search, C
i
. In this case, the user
should use the variables that define to that class V
C
i
.
4 AN APPLICATION EXAMPLE
This section show briefly the application of the search
method suggested to an example.
The user selects the class in which is interested,
rent houses, and studies the variables that define to
this class with the aim to value his ideal item, o
i
, ac-
cording to this information. For this case, the set of
variables that are used are shown in Table 1.
Table 1: Variables that define rent houses class.
Variable Type Value range Domain
Price Continuous Yes [100, 700]
Rooms Graduated Yes [1, 5]
District Ordinal Yes [1, 4]
Internet Boolean No
Street Nominal No
The definition domain of the Price variable is
shown in Fig. 2.
      

 


Figure 2: Definition of the Price. variable.
The user values each variable and associates to
each one a importance grade (see Table 2).
Table 2: Description of the user item, o
i
.
Variable Value Range Importance
Price (300,400,400,500) No Maxim (1)
Rooms 3 No High (0,75)
District [1, 3] Yes Low (0,5)
Internet Yes - High (0,75)
Street Altagracia - Low (0,5)
IMPROVING THE SEARCH AND CATALOGUING OF ITEMS IN C2C E-COMMERCE PORTALS
17

Figure 3: Portal’s screen capture.
Table 3: Items of rent houses class.
House Price Rooms District Inter. Street
C
i
v
1
C
i
v
2
C
i
v
3
C
i
v
4
C
i
v
5
C
i
e
1
550 5 4 Yes De la Rosa
a = 500
b = 550
c = 550
d = 600
e
2
200 2 4 No Altagracia
e
3
300 4 2 No De la Luz
e
4
350 3 3 Yes Pilar
a = 300
b = 350
c = 350
d = 400
e
5
500 4 1 Yes Postas
After, the user has given this information that de-
fine his ideal item (Table 2), the system proceeds to
search similar items into the class (rental houses).
If we suppose that the items which are stored in
the portal under the class C
i
(rent houses) are those
that show in Table 3. The values in cursive are ap-
proximate values, and we show how they are defined
by means of a fuzzy number.
The local distance between the item o
i
and each
item e
j
from the class C
i
(e
j
∈ C
i
), i.e. d
N
(e
j
, o
i
, v
k
C
i
)
for k = 1 to 5 are shown in Table 4.
Next, we calculate the global distance between
each e
j
and o
i
, applying Eq. 1. The following ordered
list is obtained:
Table 4: Local distances d(e
j
, o
i
, v
C
i
).
House Price Rooms District Inter. Street
e
1
0,13 0,5 0,25 0 1
e
2
0,25 0,25 0,25 1 0
e
3
0,08 0,25 0 1 1
e
4
0,01 0 0 0 1
e
5
0,08 0,25 0 0 1
e
4
(0, 51),e
5
(0, 77),e
1
(1, 14),e
2
(1, 32),e
3
(1, 52)
The application recommends to rent the item e
4
since it has the lowest global distance value (0, 51).
5 CONCLUSIONS
In this paper, we have suggested a new method based
on fuzzy logic to catalogue and search items (goods
or services) in a C2C e-commerce portals. It allow to
the buyer to carry out searches in an imprecise way.
The seller may insert items into the catalogue in fuzzy
terms.
Recently, agent technologies are being applied
to e-commerce where a personalized, continuously
running, semi-autonomous behavior is desirable
(Guttman et al., 1998; He et al., 2003). The agents
could be integrated easily in this portal. It will be our
main research for the future.
The prototype of the suggested C2C e-
ICEIS 2007 - International Conference on Enterprise Information Systems
18

commerce portal based on fuzzy logic has been
put at the following web address (see Fig. 3):
http://apps.oreto.inf-cr.uclm.es/e-alcazaba/
ACKNOWLEDGEMENTS
This work has been supported by Research Projects
e-PACTOS (ref. PAC-06-141) and SARASVATI (ref.
PBC06-0064) founded by Junta de Comunidades de
Castilla-La Mancha (JCCM).
REFERENCES
Cameron, D. (1997). The New Business Platform for the In-
ternet. Computer Technology Research Corporation,
Charleston, 1st edition.
Castro-Schez, J. J., Castro, J. L., and Zurita, J. M.
(2004a). Fuzzy repertory table: A method for acquir-
ing knowledge about input variables to machine learn-
ing algorithm. IEEE Transactions on fuzzy systems,
12(1):123–139.
Castro-Schez, J. J., Jennings, N. R., Luo, X., and Shadbolt,
N. R. (2004b). Acquiring domain knowledge for ne-
gotiating agents: a case of study. International Jour-
nal of Human-Computer Studies, 61(1):3–31.
Castro-Schez, J. J., Jimenez, L., Moreno, J., and Rodr
´
ıguez,
L. (2005). Using fuzzy repertory table-based tech-
nique for decision support. Decision Support Systems,
39(3):293–307.
Chaudhury, A. and Kuilboer, J. P. (2001). E-Business and
E-Commerce Infrastructure: Technologies Supporting
the E-Business Initiative. McGraw-Hill, Boston, 1st
edition.
Compras. Compras web site http://www.compras.com.
Ebay. Ebay web site http://www.ebay.com.
Froogle. Froogle web site froogle.google.com.
Guttman, R., Mouka, A., and Maes, P. (1998). Agent-
mediated electronic commerce: A survey. Knowledge
Engineering Review, 13(2):147–159.
He, M., Jennings, N. R., and Leung, H. (2003). On agent-
mediated electronic commerce. IEEE Transactions on
Knowledge and Data Engineering, 15(6):1345–1363.
Kowtha, N. R. and Choon, T. W. I. (2001). Determinants of
website development: a study of electronic commerce
in singapore. Information & Management, 39(3):227–
242.
Laudon, K. C. and Laudon, J. P. (2005). Management infor-
mation systems: managing the digital firm. Prentice
Hall, NJ, U.S.A, 9th edition.
Turban, E., Lee, J., King, D., and Chung, H. (2000).
Electronic Commerce: A Managerial Perspective.
Prentice-Hall, Upper Saddle River, NJ, 1st edition.
IMPROVING THE SEARCH AND CATALOGUING OF ITEMS IN C2C E-COMMERCE PORTALS
19