Transcript Interfaces Supporting Knowledge Discovery In Data (ISKDD)

Interfaces Supporting Knowledge
Discovery In Data (ISKDD)
BSE(Hons)
Name: Mark Hollands
Id: 13079042
Supervisor: Assoc. Prof. Trevor Dix
Project Aims



Interface an existing data mining system,
Snob, with the internet
Enhance the user interface
Measure the effectiveness of these
interface updates with usability testing.
Contents






Data Mining
The KDD Process
User-centered Design
Snob-Online
Results
Conclusion and Further Work
Data Mining

Data Mining
Came to popularity in the early 90s
 Driven by academic and commercial
interest
 Purpose
 Low level Data sets
 High level Knowledge Discovery

Knowledge Discovery in Databases
 Framework
to support Data Mining
 Multi-disciplinary
 Support the user’s interactions with
the DM system.
KDD Process
1.
2.
3.
4.
5.
6.
7.
Determine the problem to be solved.
Creation of the relevant dataset for mining.
Pre-processing of the dataset.
Modification of the scope of the dataset.
Data Mining
Analysis of the model against the hypotheses.
Acting upon the discovered knowledge.
User-centered Design


Software development methodology
3 Main Principals :
 Goals and subsequent actions
 Empirical measurement of usage of the
system.
 Iterative Design
Snob
Unix based data mining application
 Developed in the CSSE school
 Utilizes the Minimum Message Length
(MML) principal

Snob-Online



Data Mining Environment
Web based
KDD Focused
Snob-Online Architecture



Interactivity
Flexibility
Portability

Consistent
requirements





Users
Projects
Data
Sessions
Volatile Requirements



Commands
Output
Results
Visualisation


XML Interface
ggobi
Usability Testing




Range of Users
Specific Test Cases
Monitoring usability and knowledge
discovery
3 Stage Process



Compares basic graphical interface to
command line interface.
Adds Interpretation to the system.
Adds Visualisation to the system.
Results

Stage 1
Almost all users preferred the graphical
interface.
 Most users were capable of using the
header to assess their state within the
system.
 Users with previous Snob experience
quickly understood the system control
flow.

Results

Stage 2
Interpretation provided a small gain in
knowledge discovery for novice users.
 Experienced users of Snob saw little
knowledge discovery gains.

Results

Stage 3

Visualisation provided a large increase in
knowledge discovery
Conclusion



A KDD Focus can be used to increase the
potential usability and knowledge
discovery of a data mining system.
User-centered Design maps well to KDD
development
The portable XML interface is well suited
to the data mining domain
Further Work



Pre-processing stage
Potential extension of the system to a
generic web interface for interactive linux
applications.
Standardised XML Data Mining schema