Evaluation of Predictive Models in Institutional Research Based on Multi-Objective Optimization
DOI:
https://doi.org/10.52731/lir.v001.018Keywords:
Institutional Research, Predictive modeling, Machine learning, Multi-objective optimization, Decision-makingAbstract
In institutional research (IR), the use of predictive models based on machine learning has attracted significant attention, especially for predicting students at risk of dropping out (at-risk students) and academic success. Since various evaluation metrics for predictive models in IR can be considered, the tradeoffs among them must be taken into account in model selection. Thus, this study considers the model selection process, as a multi-objective optimization problem, and proposes a framework to visualize the results of evaluating model candidates by multiple evaluation indicators, which are important in the IR context. Specific examples of numerical experiments using actual data are also presented to summarize its effectiveness and challenges.
References
C. Brooks and C. Thompson, “Predictive Modelling in Teaching and Learning,” Handbook of Learning Analytics, pp. 61–68, SoLAR, 2017.
A. Parnell, D. Jones, A. Wesaw, and D. C. Brooks, “Institutions’ Use of Data and Analytics for Student Success,” NASPA, AIR and EDUCAUSE, 2018.
K. Deb, Multi-Objective Optimization using Evolutionary Algorithms, New York, USA: John Wiley & Sons, 2001.
N. Kondo, T. Matsuda, Y. Hayashi, H. Matsukawa, M. Tsubakimoto, Y. Watanabe, S. Tateishi, and H. Yamashita, “An Approach for Academic Success Predictive Modeling based on Multi-objective Genetic Algorithm”, International Journal of Institutional Research and Management, Vol.5, No.1, pp.31–49, 2021.
