Pertemuan 21 Software Agents for E-Commerce

1

Pertemuan 21 Software Agents for

E-Commerce

Matakuliah : M0284/Teknologi & Infrastruktur E-BusinessTahun : 2005Versi : <<versi/revisi>>

2

Learning Objectives

• Describe what software agents are• Differentiate between various classes of

software agents• Understand the use of artificial intelligence and

statistical reasoning • Describe the range of agents available to assist

in the buying process• Identify various activities in e-commerce where

software agents can be used

3

Overview

• What are software agents?• Logic of agent behavior• Types of agents• Information agents• E-Commerce agents• Mobile agents

4

What are software agents?

• Software entities– Autonomy/agency - without detailed commands– Purposeful - goal-driven– Reactive - react to changes in environment. Exhibit

intelligence– Social and Mobility skill - travel around and interact

with other agents

5

What are Software Agents?

6

Logic of Agent Behavior

• Symbolic Reasoning if <condition> then <action>

• BeyondMail from Banyan

1) IF <event='mail_receipt'> AND < email_sender=’CEO’> THEN

<save_in_folder=’Urgent’>

2) IF <save_in_folder NOT Empty> THEN <notify>

7

Logic of Agent Behavior

• Statistical Reasoning– Market Segmentation

• Clustering according to some characteristics such a buying behavior, demographic data

– Also called Collaborative filtering– Used by Amazon.com to predict books that

might prove to be your favorite

8

Logic of Agent Behavior

• Multi-attribute utility theoryused to rank-order different choices such as items to buyUtility is related to various quality, price and delivery attributesUtility numbers are calculated for various choices

Various formulas used:U(x)= log( x+ b) U(x)= a + bx + cx2 U(x) = (1/k) (1- e –kx)

where U is the utility and x is the measure of the attribute. In the case of an automobile, x could be price, quality or fuel economy.

9

Logic of Agent Behavior

• Constraint Satisfaction Approach• A way to prune a large set of choices

– Hard and Soft constraints– Options/ choices that violate hard constraints

are removed– Options left are evaluated in terms of how far

soft constraints violated

10

Logic of Agent Behavior

• Auction Protocols– English auction price start low and

move up– Dutch auction price start high and

move low– Sealed-bid auction offers in sealed

envelopes

11

Logic of Agent Behavior

Auction Engines used in e-business

12

Types of Software Agents

• Information Agents

• E-Commerce Agents

• Mobile Agents

13

Information Agents

• Information Search Agents search engines• Information filtering agents search few specific

web site and retrieve information relevant to a user

14

Information Agents

Logic of filtering agents

15

Information Agents

• Information Delivery Agents– Pull versus Push (scheduled pull). In push, the client-based

software periodically contacting the server for recent news

16

Information Agents

• Information Notification Agents

message arrives by email

17

Information Agents

• Information Reconnaissance Agents– Letizia at MIT brings to attention to users

pages of interest that are only a few links away from the current page

– The system builds up a interest profile of the user and searches neighboring pages of interest