Decision Support and Expert Systems

This week we explore decision support systems and expert systems. There is relatively less material to cover this week than we have had most other weeks. Hopefully, we will be able to look at an expert system or decision support system in more detail in the discussion forum for the week. These notes are not intended to replace the lecture (or the text).

Objectives

• list and explain the three phases in decision making
• describe the difference between structured and unstructured decision making
• describe the components that make up decision support systems and expert systems
• list examples of how decision support systems are used
• list examples of how expert systems are used
• describe typical elements of a geographic information system (GIS)
• discuss the advantages and disadvantages of automated decision making

Decision making process

• model: a representation of reality
• algorithm: a step-by-step process for solving a problem
• structured problem: a problem for which an algorithm exists which can give an optimal solution in finite time
• unstructured problem: a problem for which no algorithm exists which can give an optimal solution in finite time; this is often due to unknowns and variability
• semi-structured problem: a problem for which partial solutions may be achieved, but for which variability and unknowns may still change the expected outcomes
• parameters: categories of data considered when executing an algorithm
• Three phases of decision making process
  • intelligence gathering: relevant data and techniques and models for solving the problem are collected (possibly created)
  • design/processing: the collected data is organized, a model and technique for solving the problem is selected, the data is processed using the selected model and techniques, and potential courses of action (along with their expected results) are produced
  • choice: a course of action is selected

Decision support systems (DSS)

• DSS: an IS designed to help managers select one of many alternative solutions
• Components of a DSS
  • data management module: a database or data warehouse that holds and maintains data for the DSS; data may come from a number of sources including such systems as SCM (supply chain management) and CRM (customer relationship management)
  • model management module: contains a model or models to be used by the DSS; the model may be fixed (static), dynamic (able to change due to changes in the data), or it may be a collection of possible models from which the DSS or the user may select
  • dialog module: this is the interface between the user and the DSS; this is what the user would interact with to enter data, query the system, produce reports, etc.
  • sensitivity analysis module: this is used to determine what effect particular parameters have on the result; for example, you may be doing an analysis for a municipality where the amount of tax revenue generated is given great weight toward the outcome
• examples of DSS given in text:
  • production and retailing: used to project purchasing trends and help decide how much product to stock and where to purchase it
  • tax planning: used to make financial decisions to help reduce tax burden
  • web site planning and adjustment: used to analyze customer behavior and suggest changes to design
  • yield management: used to maximize overall revenue, often by using price discrimination
  • financial services: used to make decisions such as loan approvals
  • benefit analysis: used to help determine which package of benefits is best suited for someone's needs and budget

Expert systems (ES)

• expert systems emulate the knowledge of an expert in a narrow field/domain
• AI: Artificial Intelligence; the name given to a broad field where computer engineers and scientists try to mimic patterns of thinking and learning using computers
• knowledge base: a collection of facts and the relationships between them; much of the content may be a collection of if/then rules
• inference engine: software that combines data input by the user with a knowledge base to try to suggest a solution
• neural networks: hardware or software designed to mimic the way a brain works
• Turing Test: A test proposed by Alan Turing to determine if computers can think. To pass the Turing Test, a computer must have a dialog with humans and have the humans not be able to tell if they are talking to a computer or another person. To make the test a little more achievable, the dialog is almost always carried out as a text connection, like instant messaging or chat.
• intelligent agent: software designed to wait to perform particular operations when triggered by a specific event, such as automatically reordering an item when the stock level of that item falls below a certain value
• examples of ES given in text:
  • medical diagnosis: used to recognize patterns of diseases based on test results
  • medical management: used to suggest courses of action and help avoid bad interactions with medicines, procedures, etc.
  • credit evaluation
  • detection of insider securities trading
  • detection of common metals
  • irrigation and pest management
  • diagnosis and prediction of mechanical failure
• GDDS: Group Decision Support System

Geographic information systems (GIS)

• GIS are used together with data and rules to help solve many problems
  • optimize delivery routes
  • choose store locations
  • locate schools
  • locate public services
  • search for resources (such as oil)
  • fishing
  • agriculture

Ethics

• What kinds of decisions should computers be allowed to make?
• Who or what checks the computers to see if decisions are sensible or fair?
• How can you correct bad computer decisions?
• Who is liable for computer mistakes?

Examples

• Genetic algorithm example
• Expert system demonstration