Predicting Engineering Student Attrition Risk Using a Probabilistic Neural Network and Comparing Results with a Backpropagation Neural Network and Logistic Regression

As the need for engineers continues to increase, a growing focus has been placed on recruiting students into the field of engineering and retaining the students who select engineering as their field of study. As a result of this concentration on student retention, numerous studies have been conducted to identify, understand, and confirm relationships between student attributes and attrition. Methods of prediction have also been evaluated and compared. Utilizing the attributes found in previous studies to have correlation with student attrition, this study considers the results of three different prediction methods—logistic regression, a multi-layer perceptron artificial neural network, and a probabilistic neural network (PNN)—to predict engineering student retention at a case study university. The purpose of this study was to introduce the PNN to the study of engineering student retention prediction and compare the results of the PNN to other commonly used methods in this field of study. The accuracy, sensitivity, specificity and overall results for each method are reported, compared, and discussed as the major contribution of this paper.     

A longitudinal analysis of maternal abuse potential and developmental delays in children of adolescent mothers.

OBJECTIVE: This project was designed to examine the impact of adolescent mothers' abuse potential on the development of preschool children. The specific aims were to demonstrate relationships between maternal abuse potential and developmental problems in preschool children, to examine these relationships across time, and to determine whether maternal abuse potential predicted developmental delays after controlling for problematic parenting orientations. METHOD: Using a longitudinal design, we examined 146 first time mothers and their children. Maternal abuse potential was assessed when children were 1, 3, and 5 years old; problematic parenting orientation was assessed when the children were 6 months old; and child development (i.e., IQ, adaptive behavior, and behavior problems) was assessed at ages 3 and 5. RESULTS: Regression analyses revealed significant relationships between maternal abuse potential and a variety of developmental problems. Path analyses revealed unidirectional relationships between abuse potential predicting IQ and adaptive behaviors. Further analyses indicated that maternal abuse potential at 1 and 3 years predicted intelligence and adaptive behavior at ages 3 and 5, even when problematic parenting orientation was controlled. In contrast, children's behavioral problems at ages 3 and 5 was better accounted for by problematic parenting orientation than by abuse potential. CONCLUSIONS: The results of this study revealed that developmental delays in children of adolescent are related to abuse potential. Two pathways were found for predicting developmental delays: One pathway linked child abuse potential with IQ and adaptive functioning: the other pathway showed that problematic parenting orientation accounted for the development of emotional and behavioral problems.