From Coarse to Fine and Weak to Strong: The Impact of Scale Granularity and Rating Strength on the Ability of K-Means to Recover True Cluster Structure
The current research is undertaken to understand the degree to which K-means clustering is resilient to coarse scales and skewed distributions. Two empirical studies are conducted to evaluate how scale granularity and non-normal distributions impact cluster solutions. In both studies, important cons...
| Other Authors: | |
|---|---|
| Format: | Others |
| Language: | English English |
| Published: |
Florida State University
|
| Subjects: | |
| Online Access: | http://purl.flvc.org/fsu/fd/FSU_migr_etd-2904 |
| Summary: | The current research is undertaken to understand the degree to which K-means clustering is resilient to coarse scales and skewed distributions. Two empirical studies are conducted to evaluate how scale granularity and non-normal distributions impact cluster solutions. In both studies, important considerations in the design and testing of clustering methods are addressed. The findings demonstrate that scale granularity influences the quality of a clustering solution, whether quality is measured as cluster recovery or as local optima. However, skewed distributions did not have an impact under the conditions that were tested. Important research directions are explored. === A Dissertation submitted to the Department of Marketing in partial fulfillment of the requirements for the degree of Doctor of Philosophy. === Fall Semester, 2007. === October 19, 2007. === cluster analysis, scaling, simulation, K-means === Includes bibliographical references. === Michael J. Brusco, Professor Directing Dissertation; James G. Combs, Outside Committee Member; J. Dennis Cradit, Committee Member; Charles F. Hofacker, Committee Member. |
|---|