USING AGENTS’ ATTITUDES AND ASSESSMENTS IN
AUTOMATED FUZZY BIDDING STRATEGY
Madhu Lata Goyal and Jun Ma
Faculty of Engineering and Information Technology, University of Technology,Sydney, Australia
Keywords:
Fuzzy Sets, Agent’s Attitudes, Bidding Strategy.
Abstract:
To be successful in multi-attribute auction, agents must be capable of adapting to continuous changing bidding
price. This paper presents a novel fuzzy attitude based bidding strategy (FA-Bid), which employs dual assess-
ment technique i.e. assessment of multiple attributes of the goods as well as assessment of agents attitude
(eagerness) to procure an item in automated auction. The assessment of attributes adapts the fuzzy sets tech-
nique to handle uncertainty of the bidding process as well use heuristic rules to determine attitude of bidding
agents in simulated auctions to procure goods. The overall assessment is used to determine a price range based
on current bid, which finally selects the best one as the new bid.
1 INTRODUCTION
The emergence of electronic market places has dra-
matically increased the opportunities for the online
auctions (e.g. eBay, Amazon etc.). Intelligent agent
technology (Anthony and N.R.Jennings, 2002; Byde
et al., 2002; Greenwald and Stone, 2001; He et al.,
2003) provides a powerful mechanism to address
complex problems of dynamic pricing in automated
auctions. The agents can use different auction mech-
anisms (e.g. English, Dutch, Vickery etc.) for pro-
curement of goods or reaching agreement between
agents. The agent makes decisions on behalf of con-
sumer and endeavours to guarantee the delivery of
item according to the buyers preferences. In these
auctions buyers are faced with difficult task of de-
ciding amount to bid in order to get the desired item
matching their preferences. For this reason, the for-
malisation of bidding mechanism has received a great
deal of attention from the agent community for the
past decade. These software agents should be smart
enough to bargain a favourable deal for the user. In or-
der to be called an intelligent agent the software must
satisfy several criteria like autonomy, temporal conti-
nuity, communication and cooperation. To this end, a
number of researchers (P.Anthony and N.R.Jennings,
2003; Kowalcyzk and Bui, 2000; Luo et al., 2003; Ma
and Leung, 2007; P.Stone et al., 2001) have reported
different frameworks that help an autonomous agent
to tackle the problem of bidding in auctions. Cur-
rently, no single implementation satisfies all the cri-
teria, but there are several promising results for bar-
gaining intelligent agents.
In this paper, a fuzzy bidding strategy (FA-Bid) is
designed in an automated auction based on the dual
assessment of multiple attributes of items as well as
agents attitude on bidding item. To quantify attitudes
and to deal with uncertainty of attribute assessment
fuzzy sets technique is applied in the presented strat-
egy. The basic procedure of the strategy is shown in
Figure 1. The remainder of the paper is organized as
below. First, the detail of the presented strategy is
illustrated. Then, a simple experiment is conducted.
Related work and conclusion are discussed finally.
overall assessment
price determination
Figure 1: A Fuzzy Bidding Strategy (FA-Bid) model.
385
Lata Goyal M. and Ma J. (2009).
USING AGENTS’ ATTITUDES AND ASSESSMENTS IN AUTOMATED FUZZY BIDDING STRATEGY.
In Proceedings of the International Conference on Agents and Artificial Intelligence, pages 385-391
DOI: 10.5220/0001656503850391
Copyright
c
 SciTePress

2 A FUZZY BIDDING STRATEGY
(FA-BID)
In an automated auction, an agent’s bidding activity
is influenced mainly by two aspects, namely, 1) the
attributes of goods and 2) the agent’s attitude. Any
agent prefers to make a bid for a quality goods. Rais-
ing bids will dampen the established attitude of an
agent on the goods. All these facts require an intel-
ligent agent system, plays the role of an agent’s rep-
resentation, to adopt an appropriate bidding strategy.
Considering the existence of uncertainty in a real auc-
tion situation, this paper focuses on how to make bid
by using the agent’s personal perspective.
To make a bid for a unit of goods, the agent should
balance between his/her assessment on the goods and
his/her attitude (aspiration) to win an auction. Gen-
erally speaking, an agent has stronger eagerness to
make bid for a quality goods rather than a lower one.
The eagerness is mainly based on the assessment on
the goods. Moreover, an agent’s attitude is also influ-
enced by the bids because price is the unique factor
through which agents and an auctioneer negotiate till
make a deal. To win an auction, an agent must bal-
ance among the price (bid), assessment on the goods
and attitude to win a bid.
Roughly speaking, the bidding procedure runs as
follows:
• Firstly, evaluation on each related attributes is de-
termined.
• Then these evaluations are aggregated to form an
overall assessment on the goods.
• Next, the attitude of the agent is determined.
• Overall assessment is conducted.
• Finally, a new bid is determined.
Since in real situation uncertainty exists ubiqui-
tously in expressing assessments, eagerness as well as
their relationships with price, this paper uses fuzzy-
set-based method to process uncertainty in assess-
ment and eagerness. First of all, this paper uses a
satisfactory degree measure as the common universe
of assessment, i.e., an assessment is treated as a fuzzy
set on the satisfactory degree. Secondly, an eagerness
is expressed as a fuzzy set on the set of assessments,
i.e., the assessment set is the universe of eagerness.
In the following sections, details of the strategy is
illustrated.
2.1 Attribute Evaluation
Attribute evaluation includes two kinds of process.
The first one is individual attribute assessment, and
the second one is assessment aggregation. To imple-
ment attribute evaluation, three issues are concerned,
i.e., attribute weights (relative importance) adjust-
ment, assessment expression, and assessment aggre-
gation.
2.1.1 Weights Adjustment
Weight adjustment implements dynamically change
relative importance of multiple criteria. In a real situ-
ation an agent’s personal preference on the attributes
seldom has quickly fluctuation, i.e., the weights for
criteria is relatively stable in a long run. The ad-
justment of weights resulted from the price should be
limited to a rational range. Moreover, the adjustment
shouldn’t change the relative significance among cri-
teria other than the price because raising price alters
the relative significance of it to other criteria. In the
following, the agent’s preference is treated as an ini-
tial weight vector which is the basis of the adjustment.
To construe an initial weight vector, the Analytic Hi-
erarchy Process (AHP) method (Saaty, 1980) is ap-
plied because it is provedvalidatein practice although
it may induce inner inconsistency. Suppose the ob-
tained initial weight vector is W
(0)
.
Suppose the current bid p
c
belongs to [p
l
, p
u
] ⊆ R
where p
l
and p
u
are the lower and upper boundaries
of possible bids respectively which are determined by
the auction. Let C = {c
0
, c
1
, . . . , c
K
} be the set of
K + 1 attributes and W = {w
0
, w
1
, . . . , w
K
} is the set
of weights for attributes in C.
Because except the price agent’s assessments on
other criteria do not change, the adjustment of weight
for price should be determined first. Suppose [−δ, δ]
is the adjustable range of the weight for price and the
current net increasing of weight for price is ∆w
0
, then
the current weight vector is determined by
w
′
0
= w
0
+ ∆w
0
(1)
w
′
k
= w
k
·
1− w
′
0
1− w
0
, k = 1, 2, . . . , K. (2)
where w
k
(k = 0, 1, . . . , K) is the component of W
(0)
.
Obviously,
K
∑
k=0
w
k
= 1, (3)
and the relative significance of the criteria except for
the price will not change after this adjustment.
2.1.2 Assessment Expression
Since uncertain expressions are often used in a real
situation, this paper uses linguistic terms to express
assessments. These linguistic terms are illustrated by
fuzzy set. Moreover, the universe of these fuzzy set
ICAART 2009 - International Conference on Agents and Artificial Intelligence
386

are unified to real interval [0, 1] which means the sat-
isfactory degree of the agent to a particular attribute.
Therefore, all fuzzy sets have same universe which is
convenient for aggregating assessments.
Suppose g
k
(k = 0, 1, . . . , K) is the satisfactory de-
gree measure for attribute c
k
. Then an agent’s opin-
ion on the goods in terms of attribute c
k
is denoted
by g
k
(u) where u(∈ U
k
) is the real attribute value of
attribute c
k
and U
k
is the real universe for attribute
c
k
. For instance, departing time is an attribute for a
flight ticket. The possible departing time in a day is
from 0 : 00 to 23 : 59. For any time slot u, a client
may present a satisfactory degree such as departing at
7 : 30 is with satisfactory degree 0.9 and departing at
3 : 00 is with 0.3.
In the following, let A = {a
1
, . . . , a
n
} be the set of
used assessment terms which are fuzzy sets on sat-
isfactory degree [0, 1]. Then a numeric satisfactory
degree is transformed to a linguistic term. Continue
the above example, suppose the assessment set is as
shown in Figure 2. Notice that a
7
is with the biggest
the membership degree for 0.9, the assessment for de-
parting at 7 : 30 is a
6
by the maximum membership
degree principle. Similarly, the assessment for 0.3 is
a
2
.
0
0.6 0.8 1 0.2 0
1
0.8
0.6
0.4
0.2
0.4
a
1
a
7
a
6
a
5
a
4
a
3
a
2
˜a
Figure 2: Obtain overall assessment.
2.1.3 Assessments Aggregation
An aggregated assessment is the agent’s overall opin-
ion/preference on the goods in terms of multiple at-
tributes. Take booking a flight ticket for example, an
assessment is made on a ticket usually based on the
airlines, flight departure and arrival time, flight type,
aircraft types, seat positions, as well as price. The
change of an attribute’s value may leads to the alter-
nation of an assessment. Instinct natures of differ-
ent attributes increase the difficulty and uncertainty
for obtaining an overall assessment. Notice that an
agent’s preference on an individual attribute can be
expressed through the agent’s satisfactory degree on
that attribute. This paper uses an satisfactory degree
measure as the common universe of assessment.
Based on assessment on each individual attribute,
an overall assessment can be obtained as follows.
Suppose the individual assessments of all attributes
are v
0
, v
1
, . . ., v
K
and the weights of them are w
0
, w
1
,
. . ., w
k
respectively. Then an overall assessment is ob-
tained by
a = Agg{(v
0
, w
0
), (v
1
, w
1
), . . . , (v
K
, w
K
)} (4)
where Agg is a selected aggregation method, v
k
∈ A
(k = 0, 1, . . . , K) is the linguistic assessment on at-
tribute c
k
.
To get an overall assessment in terms of a set of
criteria, an aggregation method Agg is applied. Some
existing methods can be used here, such as OWA op-
erator (Yager, 1993; Yager, 2004), 2-tuple linguis-
tic aggregation (Delgado et al., 1999; Delgado et al.,
2001; Herrera et al., 2001), and Weighted-sum. For
convenience, we use the weighted-sum-based method
to obtain an overall assessment as follows.
First, we construct a fuzzy set ˜a on [0, 1] through
˜a(u) =
K
∑
k=0
w
k
· v
k
(u), u ∈ [0, 1], (5)
where v
k
(u) is the membership degree of u in v
k
.
Next, we calculate the distance between ˜a and a
i
∈
A by
d( ˜a, a
i
) =
Z
1
0
| ˜a− a
i
|dλ. (6)
Finally, we select the nearest term(s) a to ˜a as the
overall assessment.
For example, A has seven terms, namely, a
1
, a
2
,
···, a
7
as shownin Figure 2. Suppose ˜a is the obtained
fuzzy set. By comparing the distances between ˜a and
each element in A, we know a
6
is the nearest item to
˜a. Hence, a
6
will be taken as the overall assessment.
2.2 Attitude Estimation
Customer’s attitude is his or her willingness to bid for
a unit of goods, which is related to but not the same
as the overall assessment on the given goods. After
conducting new assessment on the goods according
to current price p
c
, estimation of agent’s attitude is
implemented. In order to do so, the relationship be-
tween attitude and assessments is required. In gen-
eral, the better the assessment on the given goods is,
the stronger the attitude of bidding for that goods will
be. However, this is by no means the unique relation-
ship between attitude and assessment. For instance,
other agents’ competitive bidding sometimes can also
cause strong willingness. In this paper, we mainly
focus on the factor of assessment and extract this re-
lationship from the agent’s transaction records.
USING AGENTS’ ATTITUDES AND ASSESSMENTS IN AUTOMATED FUZZY BIDDING STRATEGY
387

Suppose E = {e
1
, . . . , e
m
} is the set of attitude ex-
pressions, A = {a
1
, . . . , a
n
} is the set of assessments,
and T = {t
1
, . . . , t
L
} is the agent’s transaction records
such that t
i
= 1 if the client won the transaction t
i
, oth-
erwise t
i
= 0. Because in each transaction, the agent’s
assessment and attitude occur simultaneously, a set of
formal rule, denoted by R, thus can be extracted from
T such that any r ∈ R is of form
r : (a
i
⇒ e
j
, α
ij
), (7)
where a
i
∈ A, e
j
∈ E, and α
ij
is the reliability degree
obtained by
α
ij
=
|{t ∈ T|a
i
, e
j
occur in t and t = 1}|
|{t ∈ T|a
i
occurs in t and t = 1}|
. (8)
Such rule depicts the approximate degree of agent’s
attitude e
j
to which the agent can win the bid un-
der the assumption that the overall assessment is a
i
.
Furthermore, these rules can be treated as a set of
fuzzy sets on A such that the membership degree in
a fuzzy set f
j
corresponding to eagerness e
j
is α
ij
.
Obviously, f
j
is an integration of rules (a
i
⇒ e
j
, α
ij
)
(i = 1, . . . , n), which is able to be treated as an alias of
e
j
. Hence, the fuzzy set f
j
is also called attitude in
the following without other specification.
Based on the rules in R, an agent can estimate the
possible attitude of the agent when it learns the cur-
rent overall assessment. set of fuzzy sets is obtained
through the following way: suppose the overall as-
sessment is a
c
, then the attitude at the moment is de-
termined by the maximum membership degree prin-
ciple
e
c
∈ E(a
c
) = {e
j
∈ E| f
j
(a
c
) > f
i
(a
c
) if i 6= j}. (9)
Notice that such determined e
c
may not necessarily be
unique. In the following, we call E(a
c
) the candidate
attitude set under a
c
.
Once the current attitude of the agent is determine,
requirements for search new bids can then be deter-
mined. The main requirements include identifying
required overall assessment and finding the candidate
prices.
2.3 Overall Assessment
Prerequisite of overall assessment is the basic require-
ment on the goods such that the agent has the highest
possibility to win a bid under the current attitude.
To find the prerequisite of overall assessment, an
order is firstly defined in E according to the strength
of attitude. Without loss of generality, suppose e
i
< e
j
if i < j. Therefore, it is possible to select the strongest
element from E(a
c
). Then the strongest element in
E(a
c
) is chosen as the first candidate attitude to deter-
mine the prerequisite of overall assessment. From the
agent’s transaction records, a set of rules
¯
R is deter-
mined such that any ¯r ∈
¯
R is of form
¯r : (e
j
⇒ a
i
,
¯
α
ij
), (10)
where e
j
∈ E, a
i
∈ A, and
¯
α
ij
is the reliability degree
obtained by
¯
α
ij
=
|{t ∈ T|a
i
, e
j
occur in t and t = 1}|
|{t ∈ T|e
j
occurs in t and t = 1}|
. (11)
Based on the maximum membership degree prin-
ciple, a set of candidate assessment is determined
such that
A(e
c
) = {a
i
∈ A|
¯
f
i
(e
c
) >
¯
f
j
(e
c
) if i 6= j}, (12)
where
¯
f
i
is the counterpart to f
i
. Each element a in
A(e
c
) is called a candidate assessment under eager-
ness e
c
.
2.4 Agent Price Determination
An agent’s assessment demonstrates some expection
on the quality of the goods. As other criteria except
the price are seldom changeable in an auction, this is
regarded in terms of price.
Suppose U
0
= [p
l
, p
u
] is the real range of price. A
price range U(a) corresponding to a candidate assess-
ment a is a subset ofU
0
such that for any u ∈ U(a), the
assessment based on u and W is a. Notice that an as-
sessment is a fuzzy set on the satisfactory degree [0, 1]
which is the bridge between assessment and price, a
price range is determined by the following steps.
Step 1: We divide the satisfactory degree [0, 1]
into n subsets D
1
, D
2
, . . ., D
n
such that
a
i
(d) > a
j
(d) (13)
for any d ∈ D
i
and j 6= i, i.e., element in D
i
with
biggest membership degree in a
i
.
Step 2: For D
a
corresponding to a candidate as-
sessment a, we select price inU
0
such that g
a
(u) ∈ D
a
.
U
a
is called a candidate bid set. Concerning that the
satisfactory degree is continuously change with the
price, we assume that U
a
is an interval in U
0
. Hence,
let p
la
and p
ua
be the left and right boundary of U
a
.
Thus, a candidate price range for assessment a is
determined.
Suppose for any element in A(e
c
), we have ob-
tained a corresponding candidate price range. Be-
cause new bid should higher than the present price
p
c
, a candidate price set for A(e
c
) is determined by
U
A
= {p
li
|p
li
> p
c
, a
i
∈ A(e
c
)}
∪ {p
ui
|p
ui
> p
c
, a
i
∈ A(e
c
)}.
As it can be seen that the candidate price range
may not exist under some assessments, in these case,
ICAART 2009 - International Conference on Agents and Artificial Intelligence
388

a weaker attitude is selected to repeat the candidate
price determination process until a range is found or
the attitude is weaker than an acceptable level.
Suppose U
A
is a found price range, there must be
a smallest element b in it. Then b is selected as the
new bid.
3 EXPERIMENT EVALUATION
In this section, an experiment implements the fuzzy
bidding strategy in a scenario in which an agent in-
tends to book flight tickets. Six factors (as shown in
Table 1) are concerned in this situation, i.e. ticket
price (c
0
), depart time (c
1
), arrival time (c
2
), num-
ber of stops (c
3
), seat positions (c
4
), and travel season
(c
5
). The flight ticket bid for is a return ticket to des-
tination D with the following properties:
- price: $800 – $2000;
- depart time: 18:00 PM, Wednesday;
- return arrival time: 10:00 AM, Friday;
- number of stops: 1;
- seat position: window;
- travel season: April (off-peak season).
Table 1: Concerned attributes of a flight ticket.
Attributes Symb. Values range Weights
price c
0
$[800–2000] 0.4
depart time c
1
Sun. 0:00 – Sat. 24:00 0.1
arrival time c
2
Sun. 0:00 – Sat. 24:00 0.1
stops c
3
0, 1, 2, 3 0.1
seat position c
4
window, aisle, middle 0.1
flight season c
5
Jan. 01 – Dec. 31 0.2
Suppose the identified perspective of an agent is
summarized as below:
• The agent prefers to a cheaper ticket and agrees to
that the cheaper the better.
• The agent prefers to travel at the weekend rather
than at working day.
• The agent prefers to no stop travel.
• The agent prefers to aisle seat then window seat.
• The agent prefers to travel during off-peak season
rather than peak season.
• The agent thinks the flight price is the most impor-
tant factor, secondly the travel season, and other
factors are of same importance.
Based on the agent’s perspective, the agent evaluates
the ticket using seven terms (shown in Figure 3), i.e.,
very bad (a
1
), bad (a
2
), slightly bad (a
3
), acceptable
(a
4
), fairly good (a
5
), good (a
6
), and very good (a
7
).
The seven terms are expressed by fuzzy sets on the
satisfactory degree [0, 1] as below (see Figure 3):
f
a
i
= e
−162(x−(i−1)
1
6
)
2
, i = 1, . . . , 7. (14)
The assessment on each individual factor is
Attribute Assessment
c
0
(no assessment)
c
1
good (a
6
)
c
2
fairly good (a
5
)
c
3
slightly bad (a
3
)
c
4
acceptable (a
4
)
c
5
good (a
6
)
membership degree
satisfactory degree
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
a
4
a
3
a
2
a
1
a
5
a
6
a
7
Figure 3: Assessment terms.
As the ticket price is the changeable factor, the as-
sessment on it is determined dynamically. For conve-
nience, suppose the agent’s satisfactory degree mea-
sure on price is expressed by a linear function as be-
low:
g
0
(p) =
2000− p
1200
. (15)
Now assume the current price (p
c
) is $900, the
agent is required to determine a new bid in this sit-
uation.
First, the satisfactory degree of the current price
is calculated by Eq. (15), which is 0.91. Because
f
a
7
(0.91) = 0.35 and f
a
6
(0.92) = 0.82, the assess-
ment for p
c
is a
6
(good).
Next, since the price changes will affect the
weights of all factors, a new overall assessment of
the ticket is calculated. Suppose the increase of price
weight is 0.05, i.e., the current weight of price is
w
′
0
= 0.45. Then the weights of other factors are cal-
culated by Eq. (2), and they are
w
′
1
= w
′
2
= w
′
3
= w
′
4
= 0.09
w
′
5
= 0.19.
Therefore, a fuzzy set ˜a(u) is obtained ( ˜a(u) in Figure
4). Then by Eq. (6), the most nearest assessment to ˜a
USING AGENTS’ ATTITUDES AND ASSESSMENTS IN AUTOMATED FUZZY BIDDING STRATEGY
389

0
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
0.1
0.2
˜a(u)
a
6
Figure 4: Overall assessment.
Table 2: Rule set for attitude estimation.
attitude
ass. e
1
e
2
e
3
e
4
e
5
a
1
0.17 0.23 0.20 0.27 0.13
a
2
0.10 0.28 0.22 0.26 0.13
a
3
0.10 0.26 0.18 0.32 0.13
a
4
0.17 0.26 0.23 0.23 0.12
a
5
0.12 0.25 0.27 0.21 0.16
a
6
0.12 0.26 0.26 0.23 0.13
a
7
0.12 0.24 0.31 0.24 0.10
is a
6
. So the new overall assessment for the ticket is
a
6
.
Then the agent needs to estimate the agent’s atti-
tude according to this assessment. Suppose the agent
uses five terms to distinguish the attitude, i.e., none
(e
1
), slightly (e
2
), medium(e
3
), strong (e
4
), and very
strong (e
5
). In order to estimate the agent’s attitude, a
set of rules of form Eq. (7) are extracted from a his-
torical auction records, which are illustrated in Table
2 and Figure 5.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
1 2 3 4 5 6 7
e4
e2
e3
e1
e5
Figure 5: Illustration for rule set R.
By Figure 5, the agent’s attitudes at this moment
are e
2
and e
3
because they have the highest reliability.
Because e
3
is stronger than e
2
, the agent first searches
possible bids under the attitude e
3
. Based on e
3
, the
Table 3: Rule set for prerequisite of assessment identifica-
tion.
assessment
att. a
1
a
2
a
3
a
4
a
5
a
6
a
7
e
1
0.11 0.16 0.12 0.24 0.16 0.16 0.06
e
2
0.07 0.21 0.15 0.18 0.16 0.17 0.06
e
3
0.07 0.18 0.11 0.17 0.18 0.19 0.09
e
4
0.09 0.20 0.19 0.16 0.13 0.16 0.06
e
5
0.08 0.19 0.15 0.16 0.20 0.17 0.05
agent discovers that a
6
is the most preferred assess-
ment on the ticket through rules in Table 3. Hence,
it will determine a candidate price range based on the
assessment a
6
.
Based on Figure 3, the agent can divide the satis-
factory degree interval [0, 1] into seven sub-intervals.
In this figure, the interval corresponding to assess-
ment a
6
is $[900–1100] and the current price p
c
be-
longs to this interval. Hence, a new bid can be se-
lected from the interval. According to the FAB-
strategy, the smallest one greater than the p
c
(900) will
be selected. For instance, if the least increase is $50,
then the new bid b is $950.
4 CONCLUSIONS
In this paper, a novel attitude-based agent’s bidding
strategy (FA-Bid) is discussed. It was noticed that
agents, which adopt attitudes, behave more flexibly
and efficiently than agents without attitude and adapt
more easily to dynamic situations. Another unique
idea presented in this paper is that to deal quantita-
tively the imprecision or uncertainty of multiple at-
tributes of items to acquire in auctions, fuzzy set
technique is used. The fuzzy logic provides attitude
based agents provide resources in the decision mak-
ing process of bidding agent. The bidding strategy
also allows for flexible heuristics both for the over-
all gain and for individual attribute evaluations. It
also explores the relationships between evaluations of
different attributes using Analytic Hierarchy Process
method (Saaty, 1980).
There are a number of areas of further investiga-
tion. In future we would further like to explore the
development of strategies for multiple auctions. We
would also like to compare our bidding techniques
with other decision theoretic approaches to determine
the relative strengths and weaknesses of these meth-
ods. Different strategies may perform well in some
environments but may perform poorly in another. The
numbers of strategies that can be employed are end-
less and the search space is huge. To address this is-
sue, we intend to use learning techniques to obtain a
ICAART 2009 - International Conference on Agents and Artificial Intelligence
390

model of the price dynamics based on the past data
and to search for most successful strategies in prede-
fined environments in an offline fashion.
REFERENCES
Anthony, P. and N.R.Jennings (2002). Evolving bidding
strategies for multiple auctions. In Proceedings of
15th European Conference on Artificial Intelligence,
pages 187–192, Netherlands.
Byde, A., Priest, C., and N.R.Jennings (2002). Decision
procedures for multiple auctions. pages 613–620.
Delgado, M., Herrera, F., and Herrera-Viedma, E. (2001).
Combining linguistic information in a distributed in-
telligent agent model for information gathering on the
internet. In Wang, P., editor, Computing with words,
pages 251–276. John Wiley and Sons, Inc.
Delgado, M., Herrera, F., Herrera-Viedma, E., Verdegay,
J. L., and Vila, M. A. (1999). Aggregation of linguistic
information based on a symbolic approach. In Zadeh,
L. and Kacpryzk, J., editors, Computing with Words in
Information/Intelligent Systems I. Foundations, pages
428–440. Physica–Verlag.
Greenwald, A. and Stone, P. (2001). Autonomous bidding
agents in the trading agent competition. IEEE Internet
Computing, pages 52–60.
He, M., Leung, H., and Jennings, N. (2003). A fuzzy-
logic based bidding strategy for autonomous agents
in continuous double auctions. IEEE Transactions on
Knowledge and data Engineering, 15(6):1345–1363.
Herrera, F., Herrera-Viedma, E., and Chiclana, F. (2001).
Multiperson decision–making based on multiplicative
preference relations. European J. Operational Re-
search, 129:372–385.
Kowalcyzk, R. and Bui, V. (2000). On fuzzy e-negotiation
agents: Autonomous negotiation with incomplete
and imprecise information. In Proceedings Dexa e-
Negotiation Workshop.
Luo, X., Jennings, N., Shadbolt, N., Leung, H., and Lee, J.
(2003). A fuzzy constraint based model for bilateral,
multi-issue negotiation in semi-competitive environ-
ments. Artificial Intelligence, 148(1-2):53–102.
Ma, H. and Leung, H.-F. (2007). An adaptive attitude bid-
ding strategy for agents in continuous double auc-
tions. Electronic Commerce Research and Applica-
tions, 6:383–398.
P.Anthony and N.R.Jennings (2003). Developing a bid-
ding agent for multiple heterogeneous auctions. ACM
transactions on Internet Technology, 3(3):185–217.
P.Stone, Littman, M., S.Singh, and M.Kearns (2001). Attac-
2000: An adaptive autonomous bidding agent. Jour-
nal of Artificial Intelligence Research, 15:189–206.
Saaty, T. (1980). The Analytic Hierarchy Process. McGraw
Hill, NY.
Yager, R. R. (1993). Families of OWA operators. Fuzzy Sets
and Systems, 59:125–148.
Yager, R. R. (2004). OWA aggregation over a continu-
ous interval argument with applications to decision
making. IEEE Transactions on Systems, Man, and
Cybernetics–Part B: Cybernetics, 34(5):1952–1963.
USING AGENTS’ ATTITUDES AND ASSESSMENTS IN AUTOMATED FUZZY BIDDING STRATEGY
391