A neural network-based decision support system for identifying and remediating at-risk students in developmental mathematics.

摘要

Developmental mathematics courses at post-secondary institutions have high failure rates that cost institutions financially and students educationally. The resulting specific research questions of this study addressed whether an early warning decision support system based on neural networks and a supporting pedagogical initiative could reduce these failure rates. Using an experimental design, at-risk students at Fort Lewis College were randomly placed into treatment (n=22) and control (n=20) groups. The resultant neural network inputs were used to make managerial decisions regarding pedagogy for a weekly one-hour concurrent support class. Cohen's d for the treatment indicated an overall effect size of 0.43, a moderately strong effect for behavioral data. The results showed an 8% increase in student success rates between 2006 and 2007. This research contributes to the existing body of knowledge by uniquely combining neural network pruning processes to create a network that identifies at-risk students. The results also suggest that using motivational interviewing, addressing math anxiety, and tailoring the support course to the needs of the students are very beneficial for at-risk students. From a positive social perspective, beyond the potential improvement in the success rate of students in the field of mathematics, the results of this research may be regarded as a framework for additional early warning systems and means for informed pedagogical initiatives for other disciplines.