Date of Award
Thesis - Restricted
Master of Science (MS)
A generalized approach is proposed to create and explain classifications using an innovative combination of artificial neural network (ANN) and fuzzy logic techniques. The significant contribution of this approach is the incorporation of fuzzy logic techniques into the interpretation of the trained ANN. This combination produces two significant benefits not offered by the neural network alone. First, it identifies those samples from the training set which provide the best example of each category discovered by the ANN. Second, it identifies those attributes which most clearly distinguish the categories. Given this knowledge, attributes which are less significant in distinguishing categories can be pruned from the input set, producing a more sharply focused classification. The approach is demonstrated by classifying general purpose software applications, based on a description of each software application in terms of attributes which are visible to a user of that application. It is significant that the techniques presented here can derive useful categories based on rather broad, external characteristics of the software. This makes the techniques useful to users of off-the-shelf software or to developers in the early stages of program specification, when the implementation details of the software are not yet clearly understood.
Clinkenbeard, Robert A., "An Unsupervised Learning and Fuzzy Logic Approach for Software Category Identification and Capacity Planning" (1992). Master's Theses (1922-2009) Access restricted to Marquette Campus. 4412.