Decision Support and Artificial Intelligence Jack G. Zheng July 11 th 2005 MIS Chapter 4

Preview:

Citation preview

Decision SupportandArtificial Intelligence

Jack G. ZhengJuly 11th 2005

MIS Chapter 4

2

Overview

Decision supportCan computers help people to make

decisions?

Artificial intelligenceCan computers be like human to make

decisions?

3

Decision Making

4 general phases of human decision making (Simon 1977): Intelligence (diagnostic)

• finding needs and problems

Design (brainstorm)• finding solutions/choices

Choice• evaluating solutions and

pick one Implementation

• applying the solution

Figure 4.2 on Page 181

4

Types of Decision (1)

Structured decision There are specific criteria to judge and the answer is certain Example

• What final letter grade should I give to you?

Non-structured decision Criteria are not explicit, or no criteria at all Example

• How much database material should I cover in CIS2010?

Most decisions involves both parts How do I evaluate students performance in the class?

5

Types of Decision (2)

Recurring decision Happening repeatedly; decision criteria quite stable Example

• How much to spend on advertising next month?

Nonrecurring/ad hoc decision Happening infrequently; decision criteria may

change every time Example

• Should I buy out my competitor to expand my business?

Decision Support Systems

7

Decision Support Systems

Decision Support System (DSS) is a type of information system designed specifically to help people make (unstructured) decisions

8

DSS Components

Figure 4.5 on page 185

9

DSS Components (2)

Data managementStoring and maintaining data

• spreadsheet (data file)• database• data warehouse

User interface (UI) managementNice forms to get user inputVarious visualizations of analysis output

• Reports, tables, charts, graphs

10

DSS Components (3)

Model management Transforming data to decision related

information using models A model is a predefined pattern to process

data• Calculation: what-if models, goal seeking

models• Statistical models• Optimization • Comparison• Classification• Prediction• …

11

DSS Types

DSS includes many types:OLAPCollaboration systems

• GDSS (Group DSS)

GIS (Geographic Information Systems)• SDSS (Spatial DSS)

12

Geographic Information Systems

GIS is designed specifically to work with spatial information to enhance decision making

In GIS, various kind of data are visualized with geographical data (maps) Is GIS just a dynamic map system? Most data can be related to geography

• Population• Sales• Weather• Traffic• Crime• …

13

How Does GIS Work?

GIS visualizes data as layers

Attribute data

Spatial data

Output

14

GIS Sample Applications

Dynamic mapsYahoo!Maps, Google Maps and MapQuestMS Streets and Trips

City and regional planningSan Francisco Enterprise GIS

• http://www.ci.sf.ca.us/site/gis_index.asp SimCity

• An excellent game using GIS

Artificial Intelligence

"His love is real. But he is not."

16

Artificial Intelligence

AI (or intelligent systems, knowledge systems) is the technology to let computers to imitate human thinking and behavior in some way

What is intelligence? Who is intelligent?UnderstandingSolving problemsLearning

17

Where is AI Used in Computing?

Speech recognition Natural language understanding Image/vision processing Robotics Data mining (business use)

18

Business AI Applications

Most widely used AI applications/techniques in the business worldExpert systemsNeural networkGenetic algorithmIntelligent agents

19

Expert Systems

A system that applies reasoning capabilities, as a human expert does, to reach solutionsAlso called rule-based system (RBS) or

knowledge-based system (KBS)A simple example

• http://www.aiinc.ca/demos/whale.html

20

ES Components

An expert system, like any information system, consists of information, people and IT componentsDomain expertisePeople

• Domain expert• Knowledge engineer• User

IT component

Figure 4.9 on page 198

21

IT Components in ES

This is used to enter coded knowledge (rules)

Knowledge (rules) are

stored here.It stores and provides reasons to every step of reasoning.

This is the brain of the ES. It reasons by matching incoming data and stored rules to reach a solution.

22

Expert Systems in Action

1. Rules:a) If age < 25, then loan

risk is highb) If annual income <

50k, then loan risk is high

c) If loan risk is high, then refuse

4. Gives reason:Age<25 and income <50k high loan risk refuse

3. Pick rules•Data: age 21 Matching rule: rule a) Result: high risk•Data: income 40k Matching rule: rule b) Result: high risk•(New) Data: high risk Matching rule: rule c) Result: refuse loan

2. Incoming data:• Age: 21• Annual income: 40k

23

Applications of ES

ES is good for diagnostic (what’s wrong?) and prescriptive (what to do?) problemsComputer or car diagnosticSee more demos on

http://www.expertise2go.com/

Help desk Customer service Technical support

24

Evaluating Expert Systems

Benefits Reliable: reduce errors Consistent: provide consistency in decision making Reduce costs and improve productivity … (more in the book)

Difficulties and limitations Expertise is implicit: it’s difficult to explain Modeling process is complex The system cannot learn and adapt to new

situations; it has no common sense or judgment• Is it really intelligent?

25

Artificial Neural Network

ANN is a way to mimic human brain and neurons ANN can be trained to model complex problems

and recognize patterns from massive inputs

26

How does ANN Work

Depending on how the learning is done Back propagation

• Needs to be trained• It is usually used for prediction• For example, to predict transaction fraud or stock

performance

Self-organizing• Self trained• It is usually used to classify data• For example, to classify customers or web search results• A demo of SOM

27

Pros and Cons of ANN

ProsCan learn and adjustCan deal with large amount of dataAccurate and fastEmbeddable

ConsDon’t ask why – can’t explain; or very

difficult to explain

28

Genetic Algorithm

GA mimics the evolutionary, survival-of-the-fittest process to generate increasing better solutions to a problem

It is usually used when A problem does not have solution sets clearly

defined by known functions It is impossible to perform an exhaustive search No need to find the best solution

29

GA Concepts

Selection Survival of the fittest

Crossover Combining portions of good outcomes in the hope of

creating an even better outcome

Mutation Randomly trying combinations and evaluating the

success (or failure) of the outcome

30

How Does GA Work?

1. Randomly throw out a number of solutions (population) as the initial generation

2. Use the fitness function to evaluate each solution Bad solutions are dropped

• Selection New solutions are added through

• Crossover• Mutation

3. Repeat step 2 on the new population (generation) Until satisfying solutions are found

31

Applications of GA

Evaluating GA GA is good for optimization problems, when decision-making

involves hundreds or even millions of possible solutions GA does not guarantee the best answer GA may give different solutions

Some applications Finding optimal routes for traveling

• TSP (Traveling Salesman Problem) Finding optimal scheduling of labor Finding optimal stock portfolio combination Strategies for game playing (chess?)

32

Intelligent Agents

Agents are software that acts on your behalf to perform repetitive computer-related tasks Also called “bot” (?)

There are many uses of bots Gathering information (search bot) Biding (bid bot) Computer usage assistance (Office Assistant) …

33

Applications of Intelligent Agent

User agent, or personal agent – a secretary? Microsoft agent:

http://www.microsoft.com/msagent/default.asp Chatbot (messenger bot): http://www.alicebot.org/

Buyer agent or shopping bot To gather and compare information of products

and services on the web Example: http://bestwebbuys.com/

34

Applications of Intelligent Agent

Monitoring-and-surveillance agent, or predictive agent Observe, analyze and report Some applications include

• Monitoring network security• Monitoring email list service• Watching your competition, bid, …

Data-mining agent Software in a data warehouse to analyze data

35

Summary

Human decision making is a 4 step process

Computers are being designed to support decision making (DSS)

• GIS

to make decisions just like humans (A.I.)• Expert systems• Neural network• Genetic algorithm• Intelligent agents

36

Good Resources

All about DSS http://dssresources.com/

GIS Internet Guide http://www.gis.com/

Herbert Simon http://www.psychologicalscience.org/observer/0401/simon.html

IBM Deep Blue http://www.research.ibm.com/deepblue/

Intelligent Agents http://botspot.com/

Recommended