Sunteți pe pagina 1din 14

Buyer Agent Decision Process Based on

Automatic Fuzzy Rules Generation


Methods

Roi Arapoglou, Kostas Kolomvatsos, Stathes Hadjiefthymiades

Pervasive Computing Research


Group, Department of Informatics
and Telecommunications
University of Athens, Greece

WCCI – FUZZ 2010


Barcelona - Spain
Outline
 Introduction
 Market Members – Scenario
 Buyer Behavior – Decision Process
 Buyer Fuzzy Logic System
 Fuzzy Rules Generation
 Results

2
Introduction
 Intelligent Agents
 Autonomous software components
 Represent users
 Learn from their owners
 Electronic Markets
 Places where entities not known in advance can negotiate for
the exchange of products
 Fuzzy Logic
 Algebra based on fuzzy sets
 Deals with incomplete or uncertain information
 Enhance the knowledge base of agents

3
Market Members - Scenario
 Buyers
 Sellers
 Middle entities (matchmakers, brokers, market entities)
 Intelligent Agents may represent each of these entities
 Scenario
 Modeled as a finite-horizon Bargaining Game
 No knowledge about the characteristics of the opponent (i.e., the other
side) is available

4
Buyer Behavior – Decision process (1/2)
 The buyer stays in the game for a specific number of
rounds
 Profit
 A Utility Function is used
 U b  V  p , where V is the buyer valuation and p is the product
price
 The smaller the price is the greater the profit becomes
 Pricing Function
ptb  p0  V  (x  Tb1 )k , where p0 is an initial price,V is the valuation,
x is the number of the proposal, Tb is the deadline and k is a
policy factor (k>1:patient, k<1:aggressive, k=1:neutral)

5
Buyer Behavior – Decision process (2/2)
 Receives proposals and accepts or rejects them making
its own proposals
 Utilizes a reasoning mechanism based on FL
 The mechanism results the value of the Acceptance
Degree (AD)
 The reasoning mechanism is based on the following
parameters:
 Relevance factor (r)
 Price difference (d)
 Belief about the expiration of the game (b)
 Time difference (t)
 Valuation (V)
6
Buyer Fuzzy Logic System (1/2)
 Architecture

 Contains a set of Fuzzy rules


 Rules are automatically generated based on experts
dataset

7
Buyer Fuzzy Logic System (2/2)
 Advantages of the automatic Fuzzy rules generation
 Mainly, it does not require a lot of time in the developer side
 It does not require experience in FL rules definition
 It uses simple numbers representing values of basic parameters
 Fuzzy rules are automatically tuned

8
Fuzzy Rules Generation (1/2)
 Clustering techniques are used
 Algorithms:
 K-means
 Fuzzy C-means (FCM)
 Subtractive clustering
 Nearest Neighborhood Clustering (NNC)
 Every cluster corresponds to a Fuzzy rule
 Example
If x * (x1* , x *2 ,...,x *n ) is a cluster center the rule is:
IF x1 is x1* AND x 2 is x *2 ... AND x n-1 is x *n-1 THEN x n is x *n

9
Fuzzy Rules Generation (2/2)
 Additional techniques
 Learning from Examples (LFE)
 Modified Learning from Examples (MLFE)
 Templates for membership functions are defined
 Dataset
 They describe the policy that the buyer should have, concernig
the acceptance of a proposal
 108 rows of data
 Each row contains data for r, d, b, t, and V

10
Results (1/3)
 Fuzzy rule base creation time
Algorithm Rule Base creation time (ms)
Subtractive 35
FCM 2560
K-Means 25
LFE 20
MLFE 25
NNC 20

 Usage of the generated Fuzzy rule base in a BG


 We use the following parameters
Buyer Parameters Seller Parameters
Initial Price 100 MUs[1] Cost 250 MUs
Valuation 255 MUs Initial Profit 250 MUs
[1] MU = Monetary Unit

 We examine the Joint Utility in seven agreement zones


(theoretic maximum equal to 0.25)
(P*  C)  (V  P* ) (1)
JU  ,
where P* is the agreement price, C is the
(V  C)2
seller cost and V is the buyer valuation
11 (1)D. Zeng & K. Sycara, ‘Bayesian Learning in Negotiation’, International Journal of Human-Computer Studies, vol(48), no 1, 1998, pp. 125-141.
Results (2/3)
 Agreement zones
Buyer Valuation Agreement Zone
255 MUs 5 MUs
260 MUs 10 MUs
270 MUs 20 MUs
300 MUs 50 MUs
500 MUs 250 MUs
700 MUs 450 MUs
1000 MUs 750 MUs

 Numerical results
Scenario Agreement
Average JU Maximum JU Algorithm
No Zone
1 5 MUs 0.08 0.24 FCM, K-Means
2 10 MUs 0.14 0.24 FCM, K-Means
3 20 MUs 0.16 0.21 LFE
4 50 MUs 0.24 0.247 FCM, K-Means
5 250 MUs 0.238 0.24 MLFE
6 450 MUs 0.208 0.21 MLFE
7 750 MUs 0.17 0.172 MLFE

12
Results (3/3)
 Performance of algorithms in the BG
Algorithm Agreements Percentage Average JU Algorithm Average AD Value
Subtractive 92% 0.217 Subtractive 80.96
FCM 69% 0.219 FCM 68.91
K-Means 69% 0.202 K-Means 62.84
LFE 57% 0.223 LFE 72.65
MLFE 85% 0.244 MLFE 74.52
NNC 86% 0.244 NNC 76.58

13
Thank you!

http://p-comp.di.uoa.gr

14

S-ar putea să vă placă și