Clustering Human Feelings Using Neural Networks and Metric Topology for Small Sample Size

Manufacturers of consumer goods often conduct surveys to determine customer satisfaction as well as guide the design of future products. In order to make use of the information it is necessary to determine the most important underlying common factors in the survey results. However, inconsistency in data obtained from humans is often problematic for conventional linear statistics. Conducting surveys over large numbers of people, a method commonly used to counter this weakness, is not always feasible or cost effective. Therefore we have developed two methods to overcome the drawbacks of linear statistical methods: an artificial neural network modified to analyze small survey data over a small sample (10-30) of human subjects and a method of analysis based on the theory of metric topology. Work done for Johnson Controls, Inc. under the direction of James A Kleiss and Colleen Serafin, in partial fulfillment of the requirements of Michigan State University MTH 844, advised by Professor Ralph Svetic.