A neural network as an instrument of prediction

This dissertation describes the construction of a Sparse Distributed Memory neural network and its application to the prediction of student success. The architecture of the network and the details of its implementation are included. The data are taken from the records of students admitted to the University of Nebraska Medical Center College of Dentistry. The predictions of the network based on preadmissions data were compared with the known academic outcomes of the students. The network was successful in 65 to 72 percent of its predictions. This success rate is comparable to classification rates and exceeds prediction rates achieved by discriminant analysis on the same data. It is also comparable to prediction rates achieved by discriminant analysis in a similar study at a different institution. These results suggest that prediction of student success is a feasible application for neural networks based on the Sparse Distributed Memory paradigm.