Electrical & Computer Engineering, Department of


Date of this Version

Fall 11-5-2010


A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Electrical Engineering, Under the Supervision of Professor Hamid Vakilzadian. Lincoln, Nebraska: November, 2010
Copyright 2010 Panpan Hu


Recent studies have shown that Artificial Neural Networks (ANNs) are suitable for recognizing patterns in the medical area. However, no study has been done to show whether or not they are also effective in the psychological area. In this study, ANNs are developed for six psychological cases related to sociobehavioral functioning. The cases are independent living skill deficits, disorder management deficits, occupational skill deficits, social skill deficits, dysregulation of anger/aggression, and substance abuse. Two models, one based on a backpropagation algorithm and the other based on a posteriori probability approach, were developed. The models were tested using data from 118 patients in a Community Transition Program (CTP). For each case, a certain percentage of data was randomly selected for training the network, and the remaining data were used for testing the network. DESIRE was used to test the developed models. The networks using DESIRE correctly identified 61.0%, 56.8%, and 56.1% of the test cases in the dysregulation of anger/aggression, substance abuse, and social skill deficits models, respectively. The results were also compared with those obtained from the MATLAB Neural Network toolbox. While MATLAB builds the model internally without identifying the type of the model it is building, the performance results were close to those obtained by DESIRE. The neural networks for dysregulation of anger/aggression and social skill deficits were grouped into a single network, which provided 42.1% accuracy on the test data. However, the combined network of dysregulation of anger/aggression and substance abuse achieved 36.0% accuracy for the test data. Finally, the ANNs developed were used to identify the 6 problem assessment cases for the 37 untitled patients in the CTP database. The results obtained illustrate that the ANN approach can be a valuable method for mining the data for clinical assessment. While there was not enough data in the database to completely train the models, the results obtained from the limited CTP database show that ANN can be a promising method of identifying patterns of psychological problem with a high degree of accuracy. However, more data is needed in order to make a definite conclusion on its increased predictability.

Advisor: Hamid Vakilzadian