SIT717 Enterprise Business Intelligence: Data Mining Presentation

This presentation delves into various data mining approaches and the critical issue of reliability in knowledge discovery. It addresses questions such as the reliability of discovered knowledge and its effectiveness in reflecting regularities in transaction data. Key factors affecting reliability, including data quality (noise, incompleteness, and class imbalance), knowledge granularity, and the dynamic nature of information over time, are discussed. The presentation also touches upon methods for measuring reliability, such as repeating tests and assessing consistency across different test scores, along with association-based, clustering-based, and classification-based approaches. References to relevant studies are included, emphasizing the importance of dynamic knowledge discovery to ensure data reliability. Desklib provides access to similar presentations and study resources for students.

Name:
Institutional Affiliation:

 A number of data mining approaches have been available
for use.
 However, the issue of reliability becomes very critical.
 One important question is, “is discovered knowledge fully
reliable?”
 Another important aspect is if it effectively reflects the
regularities hidden in transaction data.

 Data quality is the first factor affecting reliability
 Two aspects of data quality are noise and incompleteness,
and class imbalance
 Noise and incompleteness is a common factor affecting
reliability
 Discovered knowledge could also encounter class imbalance,
which could affect reliability

 Knowledge granularity is an important factor affecting reliability
 This is especially important in retain knowledge discovery
 Example, the associations considered in market basket analysis
 Sometimes certain associations could be impractical; therefore,
affecting data reliability.

 Different information could change over time
 Patterns and data hidden in various databases may also
change over time.
 To ensure data reliability, knowledge discovery needs done in
a dynamic manner.
 Knowledge discovery may be unreliable using static way
without capturing the dynamics

 Measured through repeating test over same population
 Some studies indicate that this technique could not be relied
upon since repeated tests could change
 Consistency may be assessed using test scores from different
occasions
 If results from different tests are consistent, then discovered
knowledge is reliable

 Association-based approach
 Clustering-based approach
 Classification based approach

 Dai, H., Liu, J. N. K., & Smirnov, E. (2012). Reliable
knowledge discovery. New York: Springer.
 Holsheimer, M., & Siebes, A. P. J. M. (1994). Data mining:
The search for knowledge in databases. Amsterdam, the
Netherlands: Centrum voor Wiskunde en Informatica.