Date of Award
2008
Degree Type
Dissertation
First Advisor
Paul Mangiameli
Abstract
The first essay in this dissertation proposes a new information systems design theory. IS design theories are prescriptive, specifying testable hypotheses of what artifacts and processes lead to systems that are more successful. The new design theory proposed was derived from the analysis of a case study in which information systems were developed and used to support the strategic decisions of a manufacturer of confectionary products seeking to expand in the U.S. market. These decisions were unstructured and the decision-making process an emergent knowledge process (EKP) as defined by researchers. The new design theory is in the form a set of design principles and is different from previous EKP design theories. The second essay researches how choosing a particular decision tree algorithm--classification or regression--can lead to better results under conditions of random and systematic noise. Systematic noise in particular has not been well researched despite the fact that is frequently found in the real world. Tests of the hypothesized differences between algorithms demonstrate that they are significant and that these differences can be sizeable. From patterns found in these results, an algorithm model is developed which shows how to select the algorithm with the better performance. Data quality, a term often used synonymously with information quality, is a perennial challenge for research and practice alike. The third essay is a comprehensive analysis of the four dimensions of data quality many researchers consider the most important: accuracy, completeness, consistency, and timeliness, and the effects of varying levels of data quality along those four dimensions on the outcomes of data mining. Problem complexity is integral to data mining algorithms and accordingly this essay also proposes hypotheses for the interaction effects of data quality and problem complexity. The results shows that no only do all four dimensions of data quality have significant effects but that patterns can be found in those effects. This research also shows that there are significant interactions between problem complexity and data quality in three of the four dimensions.
Recommended Citation
Blake, Roger Holland, "Three essays on success factors for information systems" (2008). Open Access Dissertations. Paper 2238.
https://digitalcommons.uri.edu/oa_diss/2238
Terms of Use
All rights reserved under copyright.