23
1 ISAT 252:Analytical Methods IV ISAT 252:Analytical Methods IV L2: Knowledge-based systems L2: Knowledge-based systems (KBS): (KBS): Definitions and requirements Definitions and requirements overview overview KBS READING ASSIGNMENT2

ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

  • Upload
    kalei

  • View
    48

  • Download
    3

Embed Size (px)

DESCRIPTION

KBS READING ASSIGNMENT2. ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS): Definitions and requirements overview. Objectives:. Students should be able to: Define a KBS and an ES, and explain the differences Describe the human roles and resources required to build a KBS - PowerPoint PPT Presentation

Citation preview

Page 1: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

1

ISAT 252:Analytical Methods IVISAT 252:Analytical Methods IV

L2: Knowledge-based systems L2: Knowledge-based systems (KBS): (KBS):

Definitions and requirements Definitions and requirements overviewoverview

KBS READING ASSIGNMENT2

Page 2: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

2

Objectives:Students should be able to:

• Define a KBS and an ES, and explain the differences

• Describe the human roles and resources required to build a KBS

• Describe and label the architecture of a KBS and its components

• Describe and explain a KBS development methodology

• Apply all of the above to a simple credit approval example

Page 3: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

3

Artificial Intelligence Artificial Intelligence (AI)(AI)

AI is a field of study that explores how computers

can be used for tasks that require (the human

characteristics of) intelligenceintelligence, such as the ability

to reason or learn, carry on a conversation, do

more or less complicate physical task. AI includes the following branches of study:

Problem solvingProblem solving Natural LanguagesNatural Languages Artificial Neural NetworksArtificial Neural Networks RoboticsRobotics Expert/Knowledge-based SystemsExpert/Knowledge-based Systems ……

Page 4: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

4

Expert/Knowledge-Based Expert/Knowledge-Based

Systems (EKBS)Systems (EKBS)

EKBS is the branch of AI that “models” computing systems on reasoning.

It present the computer as knowledgeable or “expert” on some particular topic Higher Ed Advice Medical diagnostics

Page 5: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

5

Knowledge-Based SystemsKnowledge-Based Systems

A Knowledge-Based System (KBS) is a computer system that emulates human reasoning in a relatively narrow domain of human knowledge.

An Expert System (ES) is a KBS that emulates human reasoning that embodies “significant” expertise

Knowledge-Based Systems = Knowledge + Inference Engine

Student Build Examples: KBS for helping students select a college to

attend KBS for helping with restaurant selection KBS for helping select a “nutritious snack”

SweetTooth !

Example

Page 6: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

6

E/KBS Problem CategoriesE/KBS Problem Categories

diagnosticsdiagnostics (& repair)

determining the cause of malfunctions base on observed systems (& recommending repair)

designdesign configuring a systems components to performance goals and constraints..

planning devising a sequence of actions that will achieve some goal from certain starting conditions.

instructioninstruction detecting and correcting deficiencies in student’s understanding of some domain

monitoringmonitoring comparing observed behavior of a system to expected behavior (governing the behavior of a monitored system). Monitoring a power plant

Page 7: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

7

E/KBS Problem CategoriesE/KBS Problem Categoriescontrolcontrol governing the behavior of a (monitored) system:

corrective responses to the beer (chemical) fermentation system.

interpretation interpretation forming high-high level conclusions/recognition from descriptions

collections of data: e.g., identifying ships, faces, ...

predictions?predictions? projecting probable consequences of a given situation:

stock market variables

which horse will win the race.

selectionselection assisting in choosing the best-for-the-client item from a set of alternatives

KBS SYSTEMS BUILT WITH EXSYS http://www.exsys.com/case2.html

Page 8: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

8

KBS human resource requirements KBS human resource requirements - roles/actors- roles/actors

The following are involved (required) when building a KBS: knowledge engineer $$$ domain expert(ise) end user client

The following are often crucial when building a KBS in a business/industry environment: champion/mentor informed and supportive manager

Page 9: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

9

KBS architecture & KBS architecture & components:components:

user

KBS editor

inference engine

explanation system

generalknowledge-

basecase specific knowledge-

base

user interfacemay employ:

question & answer

menu-driven

natural language, or

GUI styles

Page 10: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

10

Major KBS Components

Knowledge Base* General knowledge Base Case Specific Knowledge Base

Inference Engine Perform the reasoning about the knowledge to

provide a solution

Explanation System Provides the user with an explanation for the

recommendation, why a given question is being asked … Cf human medical diagnostics …

Page 11: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

11

Knowledge Base ExampleKnowledge Base Example Crime Solving General Knowledge Base

General or Factual knowledge about the problemIf a person has boots then (s)he has a boot

Collections of relationships that define the knowledge about the problem

1. If the subject has a football boot then (s)he has a blunt instrument.

2. If the subject has a wooden leg then (s)he has a blunt instrument.

3. If the subject is a footballer then (s)he has football boots.4. If the subject has a blunt instrument then (s)he is suspect5. If the subject is covered with blood then (s))he is a suspect,

.6. If the subject knew the victim then (s)he is suspect.

Crime Solving Case Specific Knowledge BaseContains working knowledge such as facts, conclusions and other relevant information about the problem at hand

The victim was killed with a blunt instrument.The subject knew the victimThe subject is covered in mud.The subject is a footballer.

Page 12: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

12

1. Problem definition Shared with all software development projects ..

2. Acquisition of domain knowledge knowledge acquisition is the hardest and most

time-consuming part.

3. Representation of domain knowledge

several forms of knowledge representation are used, for example, pseudo code and decision trees. Cf: flow charts.

4. Designing a solution

prototyping involves designing for something less than the complete system. Show and tell emphasis …

5. Coding the solution

this is often the easiest part.

6. Testing and correcting the solution …

Every step of the way, this is an iterative process. Note: this was not featured in developing a VB project. …

KBS Development

Page 13: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

13

Credit authorization problemWe have a problem with credit authorization. With the growth in business that we are experiencing, we have more and more customers who don’t pay their bills in a timely fashion. And we’re losing money. Several employees try to deal with this, but they don’t all do it the same way.

Sometimes a good customer that we want to keep may be late and we want to interact with them carefully. Sometimes it’s a new account and we don’t know what we want to do. Sometimes the customer has a history of bad payment and we need to do something.

For example, Ann looks primarily at their payment history. She gives long term customers and those with good payment records priority treatment, but she gives customers who do not have long-standing familiarity and who have a bad payment history what we think of as normal treatment.

Tom, however, starts with their average account balance. If they are big spenders, he gives them priority treatment. If they are not big spenders, he gives them normal treatment. However if they are friends or relatively new customers, he puts them in a special category.

We have problems! Can you help us provide more efficient, consistent, treatment?

Problem definition and Knowledge Acquisition: Any questions, clarification?

Page 14: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

14

1. How long … leads to preferential treatment?

2. What is preferential treatment?

3. What is a good payment record?

4. What are the customer/business demographics?

5. How old is a new account?

6. What is a good customer?

7. When is a customer’s history bad?

8. What is a timely fashion?

9. How many late (delinquent) payment qualifies as a bad history?

10. What is long-term, a long term customer?

11. What is priority (normal) treatment?

12. Do you reward for early payments?

Client, User, Domain Expert?

Client: store/business; business; person in charge of billing?

We use “client” to refer to the whoever will pay the bill!

Users: customers, customer service, service reps; employees who deal with customers?

There could be multiple users all with different needs!

Domain experts, expertise: keepers of the knowledge, source to-be of the knowledge-base; service reps., billing department, highest level management, hard copy?

(Re) sources could be contradictory!

Page 15: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

15

Credit authorization problemWe have a problem with credit authorization. With the growth in business that we are experiencing, we have more and more customers who don’t pay their bills in a timely fashion. And we’re losing money. Several employees try to deal with this, but they don’t all do it the same way.

Sometimes a good customer that we want to keep may be late and we want to interact with them carefully. Sometimes it’s a new account and we don’t know what we want to do. Sometimes the customer has a history of bad payment and we need to do something.

For example, Ann looks primarily at their payment history. She gives long term customers and those with good payment records priority treatment, but she gives customers who do not have long-standing familiarity and who have a bad payment history what we think of as normal treatment.

Tom, however, starts with their average account balance. If they are big spenders, he gives them priority treatment. If they are not big spenders, he gives them normal treatment. However if they are friends or relatively new customers, he puts them in a special category.

We would like to provide uniform treatment. Can you help us? Knowledge representation

Best you can tell, how does Ann reason to assigning treatment categories, what are her “business rules”?

Page 16: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

16

Ann’s business rules?

1 If If the customer has not been assigned a billing categoryand the customer has a good payment recordThenThen assign the customer a priority billing category

2 IfIf the customer has not been assigned a billing categoryand the customer has a bad payment recordand the customer has been a customer for less than 10 years.ThenThen assign the customer a normal billing category

3 If If the customer has not been assigned a billing categoryand the customer is a friend or a relatively new customerand the customer has been a customer for over 10 yearsThenThen assign the customer a priority billing category

Page 17: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

17

Credit authorization problemWe have a problem with credit authorization. With the growth in business that we are experiencing, we have more and more customers who don’t pay their bills in a timely fashion. And we’re losing money. Several employees try to deal with this, but they don’t all do it the same way.

Sometimes a good customer that we want to keep may be late and we want to interact with them carefully. Sometimes it’s a new account and we don’t know what we want to do. Sometimes the customer has a history of bad payment and we need to do something.

For example, Ann looks primarily at their payment history. She gives long term customers and those with good payment records priority treatment, but she gives customers who do not have long-standing familiarity and who have a bad payment history what we think of as normal treatment.

Tom, however, starts with their average account balance. If they are big spenders, he gives them priority treatment. If they are not big spenders, he gives them normal treatment. However if they are friends or relatively new customers, he puts them in a special category.

We would like to provide uniform treatment. Can you help us?

Best you can tell, how does Tom reason to assigning treatment categories, what are her “business rules”?

Page 18: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

18

Tom’s rules?

1 IfIf the customer has not been assigned a billing category and and the customer’s account balance is high (big spender)ThenThen assign the customer a priority billing category

2 IfIf the customer has not been assigned a billing categoryand the customer’s account balance is not highThenThen assign the customer a normal billing category

3 If If the customer has not been assigned a billing categoryand the customer is a friend OR a relatively new customerThenThen assign the customer a priority billing category

Best you can tell, how does Tom reason to assigning treatment categories, what are his “business rules”?

Page 19: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

19

Tom’s rules?

1 IfIf the customer has not been assigned a billing category and and the customer’s account balance is high ThenThen assign the customer a priority billing category

2 IfIf the customer has not been assigned a billing categoryand the customer’s account balance is not highThenThen assign the customer a normal billing category

3 If If the customer has not been assigned a billing category and the customer is a friend OR a relatively new customerThenThen assign the customer a priority billing category

Ann’s rules?

1 If If the customer has not been assigned a billing category and the customer has a good payment recordThenThen assign the customer a priority billing category

2 IfIf the customer has not been assigned a billing category and the customer has a bad payment record and the customer has been a customer for less than 10 years.ThenThen assign the customer a normal billing category

3 If If the customer has not been assigned a billing category and the customer is a friend or a relatively new customer and the customer has been a customer for over 10 yearsThenThen assign the customer a priority billing category

How to resolve their differences?

Page 20: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

20

PSEUDO CODE RULES

1. IF The customer's income is less than 25,000.

THEN The customer's line of credit is denied.

 

2. IF The customer's income is at least 25,000.

and The customer's credit rating is excellent.

THEN The customer's line of credit is approved.

 

3. IF The customer's income is at least 25,000.

and The customer's credit rating is good.

and The customer has been in their present job less than 2.5 years.

THEN The customer's line of credit is denied.

 

4. IF The customer's income is at least 25,000.

and The customer's credit rating is good.

and The customer has been in their present job at least 2.5 years.

THEN The customer's line of credit is approved.

 

5. IF The customer's income is at least 25,000.

and The customer's credit rating is poor.

THEN The customer's line of credit is denied.

Page 21: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

21

PSEUDO CODE RULES

1. IF The customer's income is less than 25,000.

THEN The customer's line of credit is denied.

 

2. IF The customer's income is at least 25,000.

and The customer's credit rating is excellent.

THEN The customer's line of credit is approved.

 

3. IF The customer's income is at least 25,000.

and The customer's credit rating is good.

and The customer has been in their present job less than 2.5 years.

THEN The customer's line of credit is denied.

 

4. IF The customer's income is at least 25,000.

and The customer's credit rating is good.

and The customer has been in their present job at least 2.5 years.

THEN The customer's line of credit is approved.

 

5. IF The customer's income is at least 25,000.

and The customer's credit rating is poor.

THEN The customer's line of credit is denied.

EXSYS Corvid

variables

values

goal-variables

Page 22: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

22

CORVID Goals:

The customer's line of credit is

1. approved.

2. denied.

CORVID Variables and VALUES:

1. The customer's credit rating is

1. excellent.

2. good.

3. poor.

• The customer has been in their present job 1. less than 2.5 years.2. at least 2.5 years.

• The customer's income is 1. less than 25,000.2. at least 25,000.

1. IF The customer's income is less than 25,000.

THEN The customer's line of credit is denied.

 

2. IF The customer's income is at least 25,000.

and The customer's credit rating is excellent.

THEN The customer's line of credit is approved.

 

3. IF The customer's income is at least 25,000.

and The customer's credit rating is good.

and The customer has been in their present job less than 2.5 years.

THEN The customer's line of credit is denied.

 

4. IF The customer's income is at least 25,000.

and The customer's credit rating is good.

and The customer has been in their present job at least 2.5 years.

THEN The customer's line of credit is approved.

 

5. IF The customer's income is at least 25,000.

and The customer's credit rating is poor.

THEN The customer's line of credit is denied.

KBS Assistant: DICTIONARY/GLOSSARY!

Page 23: ISAT 252:Analytical Methods IV L2: Knowledge-based systems (KBS):

24

HW 2:The department of Human Resources uses a very strict policy to approve or deny hotel reservations for visiting executives from other branches of the company. Every visiting executive has a job title assigned by the company: Director, Senior Manager, and Junior Manager. When a department submits a hotel reservation to HR, it applies the following policy to approve or deny the hotel reservation:

The first step in the hotel reservation authorization process is to determine the price of the hotel. HR uses the following policy to establish the price of a hotel: If the visiting executive’s hotel reservation is for Hotel C the cost is $45. If the visiting executive’s hotel reservation is B then the cost is $75.If the visiting executive’s hotel reservation is for A then the cost is $120. Otherwise, if the visiting executive’s hotel reservation is for AA, the cost is $250. After determining the price, HR decides whether the hotel reservation is approved or not. Whenever that visiting executive is a director, the hotel reservation is approved. Whenever the visiting executive is a senior manager that works for the Accounting Department and the cost of the hotel is greater than $100, the hotel reservation is approved. If the visiting executive is a senior manager that does not work for the Accounting Department and the cost of the hotel is greater than $100, then hotel reservation is denied. Whenever the visiting executive is a senior manager and the cost of the hotel is between $50 and $100, the hotel reservation is approved. If the visiting executive is a senior manager or a junior manager and the cost of the hotel is less than $50 then hotel reservation is approved. If the visiting executive is a junior manager and the cost of the hotel is greater than $50 then hotel reservation is denied.

Our department has submitted hotel reservations for Terry Smith, Jess Brown, and Robin Spencer to HR. We have made a reservation at Hotel B for Terry, who is a Junior Manager in our branch in Staunton.. Jess has a reservation for Hotel A and he is a Senior Manager in the Accounting Department at Fairfax. Robin has a reservation for hotel C and he is a Junior Manager at Roanoke.