Presenter: Alyssa Joy Viduya

1
Evaluation of an Extensible Data Mining Framework
Using Internet Usage Data

• Presenter Alyssa Joy Viduya
• Student No 203004027
• Supervisors Dr. C. Cilliers Prof. J. Greyling
• Date 16 March 2007

2
Overview

• Background
• Problem Description
• Related Work
• Analysis
• Implementation tools
• Risk Areas and Critical Success Factors
• Project Plan and Progress
• Conclusion

3
Background

• Extensible data mining system
• Integrate new algorithms and analysis methods
• Without knowledge of internal workings system
• Without reprogramming the data mining kernel

4
Problem Description

• Situation of concern
• Terblanche (2006) proposed an extensible data
  mining framework
• Social network domain
• Not yet tested by another person
• Not yet tested on a different domain

5
Problem Description

• Problem Statement
• To evaluate extensibility of Terblanches data
  mining framework by incorporating a data mining
  algorithm and visualisation component suitable
  for internet usage data

6
Problem Description

• Scope
• Data mining algorithm for internet usage data
• Visualisation technique for internet usage data
• Incorporate chosen algorithm
• Incorporate chosen visualisation technique
• Measure extensibility using Terblanches technique

7
Problem Description

• Constraints
• Components have to be added using Terblanches
  templates

8
Related Work

• Terblanches extensible data mining framework
  architecture

Data Layer Application Layer
User Interface Layer
Output component library
Algorithm component library
Database
GUI
API
Flat files
Data mining kernel
9
Related Work

• Literature Study
• Previous study on internet usage management
• Graeme Lee Son (2004)
• Akash Singh (2005)
• Craig Oosthuizen (2005)
• Examples of visualisation techniques

10
Analysis

• Functional Requirements
• Proposed Framework
• Extensible
• Data Mining System
• Display data from database
• Pre-process (clean) data
• Display cleaned data
• Mine data
• Visualise knowledge mined

11
Analysis of New System

• Non-Functional Requirements
• Proposed Framework
• Learnability
• Data Mining System
• Utility
• Effective
• Reliability

12
Implementation Tools

• Analysis and Design Tools
• Microsoft Visio 2003
• Development Tools
• Visual Studio .NET 2005 C
• Nevron.NET charting component

13
Risk Areas and Critical Success Factors

• Risk Areas
• Suitable data mining algorithm
• Suitable visualisation technique
• Critical Success Factors
• Satisfy functional requirements
• Determine frameworks level of extensibility

14
Project Plan and Progress
Problem Definition and Requirements Analysis
Analysis and Design
Implementation
System Testing and Evaluation
02/11/2007 Submission
19/03/2007
21/05/2007
03/09/2007
In progress
Completed
15
Conclusions

• Establish level of extensibility of Terblanches
  framework
• Apply framework to another domain
• Incorporate apt data mining algorithm and
  visualisation technique
• Using Visual Studio .Net C
• Contribution
• Research paper

16
Questions