Neural networks in the decision sciences
Neural networks are finding important applications in business. The purpose of this dissertation is to investigate the potential for neural network applications in the Decision Sciences area. This research studies the capabilities of neural networks in business decisions requiring cluster analysis, in process control of complex manufacturing systems, and in business decisions that involve classification.^ Cluster analysis, the determination of natural subgroups in a data set, is an important statistical methodology that is used in many business contexts. A major problem with hierarchical clustering methods used today is the tendency for classification errors to occur when the empirical data departs from the ideal conditions of compact isolated clusters. Many empirical data sets have structural imperfections that confound the identification of clusters. We use a Self Organizing Map neural network clustering methodology and demonstrate that it is superior to the hierarchical clustering methods.^ Manufacturing process are increasingly subject to tighter control and more frequent monitoring. It is now recognized that complex interactions of auto and cross-correlation exist in data observations from manufacturing industries. New control models that capture both multivariate and time series effects are needed to effectively monitor manufacturing processes. In this manuscript, we investigate the ability of Radial Basis Function Neural Networks, to monitor and control complex manufacturing processes that exhibit both auto and cross-correlation. We demonstrate that the Radial Basis Function Network is superior to three control models recently proposed for complex manufacturing processes, Multivariate Shewhart, Multivariate EWMA, and Common and Special Cause charts.^ Classification is the assignment of objects to one of a number of predetermined groups. Classification errors result when an object is assigned to the wrong group. These classification errors result in substantial economic and social costs. Researchers and practitioners have turned to neural classification models in a quest for higher levels of classification accuracy. The purpose of this manuscript is to contrast the classification accuracy of six neural models: back-propagation, radial basis function networks, fuzzy artmap classification networks, probabilistic neural networks, learning vector quantization networks, and modular neural network. ^
Business Administration, General|Operations Research
David Allen West,
"Neural networks in the decision sciences"
Dissertations and Master's Theses (Campus Access).