University of Houston
Department of Computer Science

In partial fullfillment of the Requirements for the Degree of
Master of Science


Anita Cheeti
will defend her thesis

Decision Systems with Data Clustering and Non-Clustering


Abstract

The importance of Decision Systems in today’s world cannot be over stressed. A Decision Support System is an interactive system that assists the decision-maker in arriving at the best possible solution to the problem quickly. Our decision systems process the large amount of information available to find the events in history that are similar to a current situation. It finds the decisions and the correctness of these decisions taken in these past cases and uses the majority rule to suggest the user with a decision. Three different methods based on data clustering and non-clustering were implemented for the search for these similar events and the results of these three methods for a sample test set were compared. The Exhaustive Search Method, ESM, searches all the events present in the database. The second method, Clustered Search Method, CSM, groups the existing data into a set of clusters prior to making a decision. So when a current situation is input, the system uses the data in the neighboring clusters to reach a decision instead of the entire database. The third method, Reduced Search Method (RSM), uses the average of all the members of the nearest cluster, which is stored in the database along with the average vector of the cluster. The ESM is the most accurate of the three methods but is extremely time-consuming. The CSM was found to be much faster than the ESM but the accuracy may be lesser due to the clustering involved. The RSM is the fastest method but again as in CSM, errors of the clustering algorithm may be carried over to the system. The user can decide which method he wants to use depending on the situation. Use of the first method would be a trade off on the time for accuracy purposes and use of the second and third methods is a trade off on the accuracy for efficiency purposes.


Date: Tuesday, April 25, 2000
Time: 10:30 AM
Place: 550-PGH


Faculty, students, and the general public are invited
Thesis Advisor: Dr. Stephen Huang