Transcript Gate Assignment Display System

Artificial Intelligence (AI)
Dr. Merle P. Martin
MIS Department
CSU Sacramento
Acknowledgements
Dr. Russell Ching (MIS Dept)
Source Materiel / Graphics
 Edie Schmidt (UMS) - Graphic Design
 Prentice Hall Publishing (Permissions)
 Martin, Analysis and Design of
Business Information Systems, 1995

Agenda
Gate Assignment Problem
 Artificial Intelligence
 Expert Systems (ES)
 ES Examples

In the Airline Industry
United Airlines' GADS
(Gate Assignment Display
System)
 Trans World Airlines' GATES
(Gate Assignment and Tracking
Expert System)

Boeing 747, 387-427 capacity
Lockheed L-1011, 252 capacity
Boeing 767, 170-227 capacity
Boeing 727, 115-134 capacity
McDonnell Douglas DC-9, MD-80
73-132 capacity
Gate Assignment Problem
Gate Assignment Problem
Constraints:
 Matching size of aircraft to gate
8 different types with United
6 with TWA
 Minimizing distances between
connecting flights
 Foreign vs. domestic flight
GATES Constraints
Constraints without exceptions
 Gate size
 Constraints with exceptions
 International versus domestic
flights
 Constraints with changing tolerances
 Turn-around times

GATES Constraints
Guidelines
 Taxiway congestion
 Convenience constraints
 Time between flights
 Distance between
connecting flights

Gate Assignment
ES benefits:
 Task of scheduling gate
assignments for a month
reduced from 15 hours
to 30 seconds.
 ES can be transferred to other
airport operations, reducing
training / operating costs.
Gate Assignment
Benefits (Cont.)
 Decrease susceptibility of
schedule to moods and
whims of schedulers.
 Gate assignments can be done
on demand with little interference
to current operations.
Gate Assignment
Benefits (Cont.)
 Managers can review impact
of changes, implement changes
(i.e., what-if analysis).
 ES integrated into airlines'
major operations / scheduling
systems through direct electronic
interfaces, thus expediting
scheduling.
Artificial Intelligence (AI)
Effort to develop
computer-based systems
that behave like humans:
 learn languages
 accomplish physical tasks
 use a perceptual apparatus
 emulate human thinking
AI Branches
Natural Language
 Robotics
 Perceptive Systems
 Expert Systems
 Intelligent Machines

Human Processing
Capabilities
Induction:
 act on inconsistently
formatted data
 fill in the gaps
 CN U RD THS
 Wheel of Fortune
 Adaptiveness

Human Processing
Capabilities

Insight:
 creativity
 create alternatives
 chess game
 perspicuous grouping
Perspicuous Grouping
Recognize that we can
handle only a few alternatives
 Short Term Memory (STM)
 Miller’s 7 +/- 2 Rule
 Zero in on a few viable alternatives
 Enumerate / select best
 Satisficing, rather than optimizing
 Herbert Simon’s 1958 Chess prediction

Computer Processing
Capabilities
Handle large volume of data
 quickly
 Detect signals
where humans sense “noise”
 Tireless

Computer Capabilities
Consistent
 Objective
 no “selective perception”
 Not distracted
 Minimal “down-time”

Issue
A Stanford Research Institute
(SRI) scientist once said,
“You needn’t fear intelligent
machines. Maybe they’ll
keep us as pets.”
 Will intelligent machines
replace us?
 Why or why not?
WHAT DO YOU THINK?
What is an ES?
Feigenbaum, 1983
“intelligent computer program
using knowledge / inference procedures
to solve problems difficult enough
to require significant human expertise;
a model of the expertise of
the best practitioners”

Components of an Expert System
Knowledge
Acquisition
Facility
Knowledge
Base
Inference
Engine
Explanation
Facility
User
Interface
User
Facts and Rules
Recommended
Action
Rule Induction
Rules
Induced
From
Example
Cases
Individual
Cases
Applied to
the Rules
Case
Classified
Through
Deduction
Induction
Deduction
(Inductive Logic) (Deductive Logic)
Check Overdraft Cases
Decision
Pay or
Reject
Pay
Pay
Reject
Reject
Pay
Decision Attributes
Overdraft
for Single
Type of
Credit or Multiple
Account Rating Checks
Regular
Good
Multiple
Student Unknown Single
Student
Poor
Single
Student
Good
Multiple
Student
Good
Single
Check Overdraft Cases (Cont.)
Decision
Pay or
Reject
Pay
Pay
Reject
Reject
Reject
Decision Attributes
Overdraft
for Single
Type of
Credit or Multiple
Account Rating Checks
Regular Unknown Multiple
Regular
Good
Single
Regular
Poor
Single
Student Unknown Multiple
Regular Unknown Multiple
Pay or Reject?
Pay or
Reject
?
Overdraft
for Single
Type of Credit or Multiple
Account Rating Checks
Regular Unknown Single
Bank Overdraft
Application
340 Cases of
check overdrafts
 Classification Variable:
 Check unpaid(0) or paid (1)

ID3 DECISION TREE
CR *DIFF<6.5
176
Yes
No
130
CR*DIFF<5.5
Yes
CR *DIFF<.035
60
125
No
59
57
50
1
DIFF<5.55
ACT*DIFF<.175
2
1
0
1
Pay
48
0
Reject
2
0
Reject
5
56
3
54
Pay
2
1
Reject
DIFF<42.2
0
53
Pay
9
56
14
2
Reject
1
15
4
1
0
0
Reject
Reject
2 ACT*DIFF<3
2
0
1
Pay
Yes
DIFF<9.4
15
4
DIFF<40.3
1
68
COV*DIFF<1.5
DIFF<1.65
DIFF<10.5
0
15
Pay
1
2
Pay
116
5 No
ACT*DIFF
<19.6
32
1
32
0
Reject
DIFF<20.5
101
1
69
0
Reject
0
1
Pay
Overall Classification
Rate: 97.7%
Reasons For Using ES
Consistent
 Never gets bored / overwhelmed
 Replace absent, scarce experts
 Quick response time

ES Reasons
Reduced down-time
 Cheaper than experts
 Integration of multi-expert opinions
 Eliminate routine / unsatisfactory
jobs for people

ES Limitations
High development cost
 Limited to relatively simple
problems
 operational mgmt level
 Can be difficult to use
 Can be difficult to maintain

When to Use ES
High potential payoff
OR
 Reduced risk
 Need to replace experts
 Campbell’s Soup

When to Use ES
Need more consistency
than humans
 Expertise needed
at various locations
at same time
 Hostile environment
dangerous to human health

ES Versus DSS

Problem Structure:
 ES: structured problems
 clear
 consistent
 unambiguous
 DSS: semi-structured problems
ES Versus DSS
Quantification:
 DSS: quantitative
 ES: non-mathematical
reasoning
IF A BUT NOT B, THEN Z
 Purpose:
 DSS: aid manager
 ES: replace manager

Issue
Does your company use
Expert Systems (ES)?
 How do they?
 How might they?
WHAT ARE YOUR
EXPERIENCES?
MYACIN
Diagnose patient
symptoms (triage)
 free doctors for
high-level tasks
 Panel of doctors
 diagnose sets of symptoms
 determine causes
 62% accuracy

MYACIN
Built ES with rules
based on panel consensus
 68% accuracy
 Why better than doctors?
 Heuristics

Stock Market ES
Reported by Chandler, 1988
 Expert in stock market analysis
 15 years experience
 published newsletter
 Asked him to identify data
used to make recommendations

Stock Market ES
50 data elements identified
 Reduced to 30
 redundancy
 not really used
 undependable
 Predicted for 6 months of data
whether stock value would increase,
decrease, or stay the same

Stock Market ES
Rule-based ES built
 Discovered that only
15 data elements came into play
 Refined the ES model
 Results were better than expert
WHY?

USA Expert Systems
Manufacturing Planning:
HICLASS - Hughes
(process plans, manufacturing instructions)
CUTTECH - METCUT
(plans for machining operations)
XPSE-E - CAM-I
(plans for part fabrication)
USA Expert Systems
Manufacturing Control:
IMACS - DEC
(plans for computer hardware fabrication
and assembly)
IFES - Hughes
(models dynamic flow of factory information)
USA Expert Systems
Factory Automation:
Move - Industrial Technology
Institute (material handling)
Dispatcher - Carnegie Group, Inc.
(materials handling system)
GMR - GM Corp.
(flexible automation assembly system)
FMS/CML - Westinghouse
(simulation for FMS design, planning, control)
Issue
“Expert systems are
dangerous. People are
likely to be dependent on
them rather than think
for themselves.”
WHAT DO YOU THINK?
Points to Remember
What is AI?
 What is an ES?
 When to use an ES
 Differences between
ES and DSS
 ES examples
