EFFICIENT MULTIPLE OBJECTIVE NEURAL NETWORK MAPPING OF STATE-WIDE HIGH SCHOOL ACHIEVEMENT
Nina Kajiji, Gordon H. Dash, Jr.
This paper focuses on the use of artificial intelligence to improve the accuracy of estimated metrics that are used in public policy planning to assess public school achievement in mathematics and the English arts. In this study we implement a truncated 2nd order translog production function in the form of a double-log model specification to explain variation in two Rhode Island state high school achievement indexes. The mathematics achievement index (MAI) and the state-wide three year English language arts index (ELAI). The results obtained by OLS estimation are compared to those generated by the application of a non-parametric multiple objective radial basis function (RBF) artificial neural network (ANN). The solution differences across the alternative model solutions produce a sharp focus on the need to disaggregate policy initiatives on the modeled prediction factors. For example, both models corroborate a long-held conventional wisdom by reporting elasticity estimates that infer a negative effect on mathematics achievement with percentage increase in the non-white population in the school system. However, the nonlinear mapping objective of the ANN amplifies this result to show that it is the increase in the interaction between the percentage of the non-white student population and those students who are eligible for a free or reduced lunch that is most responsible for the observed reduction in MAI scoring performance. Additional new insights are produced across other traditional explanatory factors.
Lecture Notes in Management Science (2011) Vol. 3: 57-72
3rd International Conference on Applied Operational Research, Proceedings
© Tadbir Operational Research Group Ltd. All rights reserved.
ISSN 2008-0050 (Print)
ISSN 1927-0097 (Online)
· Production Function Methods
· The Data
· Modeling Results
· Summary And Conclusions