Data Mining Tools and Techniques - Case Based Reasoning

Data Mining Tools

Data Mining Techniques

The major data mining techniques are:

1. Classification - Assigns data items to predefined categories (e.g., decision trees, naive Bayes)
2. Clustering - Groups similar data items without predefined labels (e.g., k-means)
3. Association Rule Mining - Finds relationships among variables (e.g., Apriori algorithm)
4. Regression - Predicts continuous values based on input variables
5. Neural Networks - Models complex patterns using interconnected nodes
6. Case Based Reasoning (CBR) - Solves new problems based on past similar cases
7. Genetic Algorithms - Uses evolutionary strategies for optimization

Case Based Reasoning (CBR)

Definition

Case Based Reasoning is a problem-solving technique that reuses knowledge and experience from previously solved similar problems (cases) to solve new problems. Instead of deriving solutions from scratch, CBR retrieves the most similar past case and adapts its solution.

The CBR Cycle (4 R's)

New Problem
     |
     v
[1. RETRIEVE] --> Find most similar past case(s) from case base
     |
     v
[2. REUSE] --> Adapt the retrieved solution to fit the new problem
     |
     v
[3. REVISE] --> Test and repair the solution if needed
     |
     v
[4. RETAIN] --> Store the new case and solution in the case base
     |
     v
Solved Problem (Case Base Updated)

Steps Explained

Key Components of CBR

• Case Base - A repository/database of past cases (problem + solution pairs)
• Case Representation - How cases are stored (feature vectors, frames, etc.)
• Similarity Measure - A function to measure how close a new problem is to stored cases (e.g., Euclidean distance, weighted features)
• Adaptation Mechanism - Rules or methods to modify retrieved solutions

Example

Past Case: A patient with fever, cough, and fatigue was diagnosed with flu and treated with rest and antiviral drugs.

New Case: A new patient has fever, cough, and body ache.

CBR Action: Retrieve the flu case, reuse the treatment, revise slightly for body ache (add pain relief), and retain the new case.

Advantages of CBR

• Reduces effort by reusing past knowledge
• Learns incrementally as new cases are added
• Works well even with incomplete domain knowledge
• Useful in medical diagnosis, legal reasoning, help-desk systems

Disadvantages of CBR

• Performance depends on the quality and size of the case base
• Similarity measurement can be complex
• Storage requirements grow as cases accumulate

Note: CBR is widely used in expert systems, medical diagnosis, customer support, and fault diagnosis applications.

Fuzzy Rule and Fuzzy Reasoning in Expert Systems

Fuzzy Rule

A fuzzy rule is an IF-THEN rule that uses fuzzy sets and linguistic variables instead of crisp (precise) values. It captures human-like reasoning by allowing partial truth values between 0 and 1.

General Form:

IF <fuzzy antecedent> THEN <fuzzy consequent>

Example:

IF temperature is HIGH THEN fan speed is FAST
IF age is YOUNG AND income is HIGH THEN loan risk is LOW

Here, "HIGH", "FAST", "YOUNG", "LOW" are fuzzy linguistic terms defined by membership functions, not exact numbers.

Fuzzy Reasoning in Expert Systems

Fuzzy reasoning (also called approximate reasoning or fuzzy inference) is the process of drawing conclusions from fuzzy rules and fuzzy inputs. It is done through the following steps:

Step 1: Fuzzification

• Convert crisp (real-world) input values into fuzzy values using membership functions.
• Example: Temperature = 75°C is mapped to membership degree: μ(HIGH) = 0.7, μ(MEDIUM) = 0.3

Step 2: Rule Evaluation (Applying Fuzzy Rules)

• Each fuzzy rule in the knowledge base is evaluated.
• For rules with multiple antecedents:
  • AND operation: take the minimum of membership values
  • OR operation: take the maximum of membership values
• Example:
  View full

  IF temperature is HIGH (0.7) AND pressure is LOW (0.4)
  THEN output membership = min(0.7, 0.4) = 0.4
  

Step 3: Aggregation

• Combine the outputs of all fired rules into a single fuzzy output set.
• Typically done using the maximum operator across all rule outputs.

Step 4: Defuzzification

• Convert the aggregated fuzzy output back into a single crisp value for action.
• Common method: Centroid (Center of Area) method $$z^* = \frac{\sum \mu(z) \cdot z}{\sum \mu(z)}$$

Role in Expert Systems

Summary

Fuzzy reasoning allows expert systems to handle uncertainty and vagueness in real-world problems by using linguistic rules and partial truth values, making them more flexible and human-like compared to classical rule-based systems.

Types of Persons Interacting with Expert Systems

Introduction

An expert system is an AI-based system that mimics the decision-making ability of a human expert. Several types of persons interact with an expert system, each having a distinct role and purpose.

Types of Persons Interacting with Expert Systems

1. Domain Expert

• The domain expert is a highly skilled and knowledgeable person in a specific field (e.g., medicine, engineering, finance).
• They provide their knowledge, experience, and problem-solving strategies to build the knowledge base.
• They work closely with the knowledge engineer to encode their expertise into the system.
• Example: A medical doctor providing diagnostic rules for a medical expert system.

2. Knowledge Engineer

• The knowledge engineer is responsible for extracting, organizing, and representing the expert's knowledge in a form the system can use.
• They design the knowledge base and inference rules.
• They act as a bridge between the domain expert and the expert system.
• They use knowledge acquisition techniques such as interviews, observation, and protocol analysis.

3. End User (Naive User)

• The end user is the person who actually uses the expert system to solve problems or get advice.
• They may have little or no technical knowledge about AI or the underlying system.
• They interact through a user-friendly interface, providing inputs and receiving recommendations or explanations.
• Example: A farmer using an agricultural expert system to identify crop diseases.

4. System Builder / Knowledge Base Builder

• This person is responsible for constructing and maintaining the expert system using expert system shells or programming tools.
• They handle the technical aspects such as coding, integration, and updating the system.
• Sometimes the role overlaps with the knowledge engineer.

5. System Administrator / Manager

• The system administrator oversees the overall operation and maintenance of the expert system.
• They ensure the system is up-to-date, accurate, and functioning correctly.
• They manage access control, system updates, and performance monitoring.

Summary Table

Conclusion

Each type of person plays a critical and complementary role in the development, maintenance, and use of an expert system. Effective collaboration among all these persons is essential for building a successful and reliable expert system.

Short Notes

a. ROMC Design Approach

ROMC stands for Representations, Operations, Memory Aids, and Control Mechanisms. It is a systematic framework used for designing Decision Support Systems (DSS). It provides a structured way to think about the components needed to support decision-making.

Components:

Significance:

• Ensures the DSS is user-centered and supports the full decision-making process.
• Helps designers identify what the system must do before deciding how to build it.
• Bridges the gap between user needs and system functionality.

b. Group Decision Support System (GDSS)

A Group Decision Support System (GDSS) is an interactive, computer-based system designed to facilitate the solution of unstructured or semi-structured problems by a group of decision makers working together.

Key Characteristics:

• Supports multiple users working simultaneously or asynchronously.
• Designed specifically for group interaction and collaboration.
• Aims to improve the quality and efficiency of group decisions.
• Reduces negative group behaviors such as groupthink, domination, and social pressure.

Components of GDSS:

1. Hardware - Networked computers, display screens, communication devices.
2. Software - Tools for voting, brainstorming, ranking, and modeling.
3. People - Participants, a facilitator who guides the session.
4. Procedures - Rules and methods for conducting group sessions.

Features:

• Anonymity - Members can contribute ideas without fear of judgment.
• Parallel Communication - All members can input ideas simultaneously.
• Automated Record Keeping - All contributions are automatically saved.
• Voting and Ranking Tools - Help reach consensus efficiently.

Types of GDSS Meetings:

• Same place, same time (Decision Room)
• Different place, same time (Teleconferencing)
• Different place, different time (Asynchronous collaboration)

Benefits:

• Increases participation and idea generation.
• Reduces meeting time and cost.
• Improves decision quality through structured processes.
• Encourages equal contribution from all members.