Agents for e-Commerce

8/7/2019 Agents for e-Commerce

http://slidepdf.com/reader/full/agents-for-e-commerce 1/6

Agents for e-CommerceJane Hsu

Motivation for Agents in E-Commerce

p Task-delegation

p Personalized

p Adaptive

p Continuously running

p Semi-autonomous

3

CBB Model (Pattie Maes, et al ) Classification of Agents: Market View

processes of sales ⇒ categories for agents:

p Demand Identificationp Awareness of the need to buy

p Product Brokeringp What to buy

p Merchant Brokeringp Who to buy from

p Negotiationp How much to pay

p Purchase and Deliveryp Payment and delivery options

p Product Service and Evaluationp Service reminder and tracking

E-Commerce Agents

p Personal shopping assistants

n Price comparison

n Compatibility

n Purchase/warrantee information

p Distributed negotiation agents

p Auction Bots

p Stock Bots

p Recommendation and notification

p Agent-mediated electronic commerce

Agent-Mediated E-Commerce [MIT]

p C2C smart classified ads

p Merchant agents

n Integrative negotiation capabilities

p Expertise brokering

p Distributed reputation facilities



Agents as Mediators in E-Commerce

Y

Personal

Logic

Y

Auction Bot

Y

Firefly

Y

Bargain

Finder

6. Product

Service and

Evaluation

5. Purchase and

Delivery

YY4. Negotiation

YY3. Merchant

Brokering

YY2. ProductBrokering

1. Need

Identification

eBayKasbahExcite

Jango

Overview

p 1st generation agents

n Filter information

n Match people w/similar interestsn Automate repetitive behavior

p 2nd generation

n E-commerce ==> revolutionizep business-to-business

p business-to-consumer

p consumer-to-consumer

Price-Comparison Shopping Agents

p BargainFinder was the first shopping agent foron-line price comparisons.

p Given a specific music CD, BargainFinderrequests its price from each of nine differentmerchant Web sites using the same request asfrom a Web browser. BargainFinder then presentsits results to the consumer.

p Like most of the first generation of e-commercesystems, BargainFinder do not exist anymore.However it offers valuable insights into the issuesinvolved in product comparisons in the onlineworld.

p Limited to comparing merchants offering only onprice instead of their full range of value

Excite's Jango

p Jango is similar to BargainFinder but with moreproduct features to search across and moreshopping categories.n Help user decide what to buy.n Finding specs and reviews of products.n Make recommendations.n Comparison shopping for best buy.n Monitoring “what’s new” lists.n Watching for special offers & discounts.

p Jango solves the merchant blocking issue byhaving the product requests originating fromeach consumer's Web browser instead of acentralised site as in BargainFinder appear asrequests from real customers



Example: MySimon

p Comparison shopping agentn User enters or chooses

description of what they arelooking for.

n mySimon presents matchesfrom merchant databases.

p Agent learns to searchn Supposedly, agent learns how

to retrieve desired informationfrom merchant websites.

n Non-programmers interactwith system to teach it how toretrieve prices from new sites.



Aristocart by Kanndu

p Conducting searches according to the criteria youspecify.

p Automatically running price comparisons on the

items you select, in order to find the best price.p Constantly monitoring the Web to see whether

prices have changed.p Enabling you to make all your purchases,

whether from one online store or many, using asingle, two-click checkout procedure.

p Download Aristocart

20

User

Information Agent

Internet

維科 SERVER

華南銀行 SERVER

Request =``java’’

Queryplanner

planexecutor

URL request

URL request

Extractor華南銀行

ExtractorWiley

Extractor維科

Thanks!

Query plan:define orders and steps toreply user’s request

Domain

model

Define the domain:

objects and theirrelations

Sourcemodel

Configuration file:describe connected

Web sites

Amazon?

Wiley SERVER

Extracted data

Extracted dataConfigureAMAZON?

ExtractorAMAZON?

Connectinga new Web site!

Recommender Systems

p Content-based filteringn Collects information from various sourcesn Synthesizes information

p Collaborative filteringn Use information about other customers to recommend

p Constraint-based filteringn Special case of content-based

n Optimization problem within constraints

Firefly

p Firefly services help consumers find products.

p Instead of filtering products based on features,Firefly recommends products via a word of mouthrecommendation mechanism called automatedcollaborative filtering (ACF).

p ACF first compares a shopper's product ratingswith those of other shoppers. After identifyingthe shopper's nearest neighbours (i.e., userswith similar tastes), ACF recommends productsthat they rated highly.

p Essentially, Firefly uses the opinions of like-minded people to offer recommendations.

p Firefly was acquired by Microsoft in 1998.



Basic calculations in collaborative filtering

a – customer 1’s preference vector

b – customer 2’s preference vector

Collaborative Filtering Agents As Mediators

Role

Buyer Seller

*Mediator Wrapper

*consults queries

Recommender Agents

Buyer Recommender

Buyer

Buyer

User

User

User

User profilezip, age, genderinterests

constraints

Catalog ofitemsitem 1

item 2...

Software Design as Problem Solving

DomainProblem

EndUser

SoftwareDesigner

The Soloist Model

programmer Computer

The Conductor Model

programmer

Computer

Computer

Computer



The Manager Model

EndUser

Agent

InterfaceAgent

Agent

Agent

Agent-Oriented Design

p A collection of task-specific agents

p Multi-agent architecture

p Agent communication

n KQML

n Eureka (KQML/ContractNet Protocols)

Agent Design Requirement

p Perceptions

p Actions

p Architecture

n Reactive

n Deliberative

n ??

p Functions performed by the agent

p A sample scenario of the agent in action

Agents: Shopping Domain

p Airfare data collection agent

p Price mining agent

p Ticket purchase agent

p Travel planning agent

p Recommender agent

p Product alert agent

To Buy or Not to Buy?

p Question: Should I buy the ticket now?

n To learn the behavior of airfares over time

p Collect airfare data

p Induce airline pricing model

Airline BPricing Model

Travel Web

Airline A

Pricing Model

Airfare Data

ShoppingAgent

DestinationConstraintsPreferences

Comparison Shopping

p ShotBot

p Price Mining

n Rule learning

n Q-learning

n Moving average models

n Combination of methods

Agents for e-Commerce

Documents