Transcript Intelligent Systems Over the Internet

Intelligent Systems Over the
Internet
By
Dr.S.Sridhar,Ph.D.,
RACI(Paris),RZFM(Germany),RMR(USA),RIEEEProc.
email : [email protected]
web-site : http://drsridhar.tripod.com
Learning Objectives
• Understand intelligent systems
operating across the Internet.
• Examine the concept of intelligent
agents.
• Learn intelligent agent applications.
• Explore the concept of Web-based
semantic knowledge.
• Understand recommendation
systems.
• Design recommendation systems.
Spartan Uses Intelligent
Systems to Find the Right
Person and Reduce
Turnover
Vignette
• Supermarket chains experience over
100% turnover
• Employee replacement expensive
• Front-end positions critical in terms of
customer relationships
• Spartan employed automated hiring
system
• Analyze applicant profile
• Selects candidates from huge applicant
pool
• Reduced turnover rate to 59%
• Increased operational efficiency
• Integrated with other systems
Intelligent Systems
• Programs with tasks automated
according to rules and inference
mechanisms
• Web used as delivery platform
• May include semantic information
• Semantic Web
• Generally perform specific tasks
• Information agents
• Monitoring agents
• Recommendation agents
Intelligent Agents
• Program that helps user
perform routine tasks
• Software agents, wizards, demons,
bots
• Degree of independence or
autonomy
• Three functions
• Perception of dynamic conditions
• Actions that affect environment
Intelligence Levels
• Wooldridge
• Reactivity to changes in environment
• Ability to choose response
• Capability of interaction with other agents
• Lee
• Level 0
• Retrieve documents from URLs specified by user
• Level 1
• User-initiated search for relevant pages
• Level 2
• Maintain user profiles
• Notify users when relevant materials located
• Level 3
• Learning and deductive reasoning component to
assist user in expressing queries
Components
• Owner
• User name, parent process name, or master agent
name
• Author
• Development owner, service, or master agent name
• Account
• Anchor to owner’s account
• Goals and metrics
• Determines task’s point of completion and value of
results
• Subject Description
• Description of goal’s attributes
• Creation and Duration
• Request and response date
• Background information
• Intelligent subsystem
• Can provide several of the above characteristics
Agents
•
•
•
•
•
•
•
•
•
•
•
Can act on own or be empowered
Can make some decisions
Can decide when to initiate actions
Unscripted actions
Designed to interact with other agents,
programs, or humans
Automates repetitive, narrowly defined
tasks
Continuously running process
Must be believable
Should be transparent
Should work on a variety of machines
May be capable of learning
Successful Intelligent
Agents
• Decision support systems
• Employee empowerment for
customer service
• Automation of routine tasks
• Search and retrieval of data
• Expert models
• Mundane personal activity
Classifications
•Franklin and
Graesser’s
autonomous
agents
• Organization agents
• Task execution for processes or applications
• Personal agents
• Perform tasks for users
• Private or public agents
• Used by single user or many
• Software or intelligent agents
• Ability to learn
Characteristics
• Agency
• Degree of measurable autonomy
• Ability to run asynchronously
• Intelligence
• Degree of reasoning and learned
behavior
• Mobility
• Degree to which agents move through
networks and transmit and receive data
• Mobile agents
• Nonmobile are two dimensional
• Mobile are three dimensional
Web Based Software
Agents
• E-mail/Mailbot agents
• Softbots:
• Agents offering assistance with Web
browsing
• Assistance with frequently asked
questions
• Search engines
• Metasearch engines
• Network agents
•
•
•
•
Monitor
Diagnose problems
Security
Resource management
E-commerce Agents
•
•
•
•
•
•
•
•
•
•
Identify needs
Search for product
Find best bargain
Negotiate price
Arrangement of payment
Arrange delivery
After sales service
Advertisement
Payment support
Fraud detection
Other Agents
• Computer interfaces
• Agents to facilitate learning
• Speech agents
• Intelligent tutoring
• Support for activities along supply chain
• Administrative office management
•
•
•
•
• Workflow, computer-telephone integration
Web mining for information
Monitoring for alerts
Collaboration among agents
Mobile commerce using WAP-based
services
DSS Agents
• Agent types
• Data monitoring, data gathering,
modeling, domain management, learning
preferences
• Holsapple and Whinston
• Map types against
− Characteristics
– Homeostatic goals, persistence, reactivity
− Reference points
– Client, task,domain
• Hess
• Map types against
− Components
– data., modeling, user interface
Multi-agent Systems
• Multiple software agents used to
perform tasks
• Multiple designers
• Agents work toward different goals
• Can cooperate or compete
• Distributed artificial intelligence
•
•
•
•
Single designer
Decomposes tasks into subtasks
Distributed problem solving
Single goal
Semantic Web
• Content presentation
• Organization standard
• Enables access to Web-based
knowledge
• Allows Web-based collaboration and
cooperation
• Technologies
• XML
• Scripting language employing user defined
tags
• Web services
• XML-based technologies comprised of four
layers
− Transport, XML messaging, service
Components of
Semantic Web
• Resource Description Framework
data model
• Relate Uniform Resource Identifiers to
each other
• Point to Web resources
• Language with defined semantics
• Standardized terminologies for
knowledge domain
• Service logic establishes rules
governing use
• Proof
• Trust
Advantages and
Limitations
• Advantages:
• Easy to understand
• Systems and
modules easily
integrated
• Saves development
time and expense
• Allows for
incremental and
rapid development
• Updates
automatically
• Resources reuse
• Limitations:
• Oversimplified
graphical
representation
• Needs additional
tools
• Incorrect definitions
• Information may be
incorrect or
inconsistent
• Security
Recommendation
Systems
• Personalized
• Collect and analyze each user’s
information and needs
• Profile generation and maintenance
−
−
−
−
−
Profiling method determination
Initial profile generation
Data processing for pattern recognition
Feedback collection
Analyze feedback and adapt
• Profile exploitation and recommendation
− Identify useful information
− Compare user profile to new items
− Locate similar users, create neighborhood,
make prediction
Recommendation
Systems
• Collaborative filtering
• Market segmentation used to predict
preferences
• Compares individual to population in order to
locate similar users
• Similarity index metrics
• Infer interests
• Predicts preferences based on weighted sums
• Content-based filtering
• Recommendations-based on similarities between
products
• Attribute based
• Works with small base of data
• Neglects aesthetic aspects of products
Management Issues
•
•
•
•
•
•
•
•
Expense
Security
Systems integration and flexibility
Hardware and software requirements
Agent accuracy
Agent learning
Invasion of privacy
Competitive intelligence and industrial
intelligence
• Other ethical issues
• Heightened expectations
• Systems acceptance