| The objective of this research is to discover temporal patterns, which represent groups of patients who have had a similar experience in course-of-disease, in a database of patients who all have the same catastrophic or chronic illness. The Event Set Sequence approach to this pattern discovery problem is proposed and implemented in the Temporal Pattern Discovery System (TEMPADIS). The entire process of knowledge discovery and data mining is investigated as it applies in this domain.; The most important contribution of this work is the view, which has not previously been demonstrated, into the mass of data collected on these patients. The fact that this view can be obtained computationally and that it reveals specific groups of patients for further study is unprecedented. Further, solutions to various barriers to the discovery process are presented.; In the data preparation phases, the issues of data comparability, missing data, and missing knowledge are addressed. In the data mining phase, TEMPADIS implements the Event Set Sequence approach and an inexact matching scheme to address issues of computational complexity and the sparseness of available data for use in discovery. An evaluation of the TEMPADIS system reveals many areas for future work. |
