METHODS OF MODELING AND FORECASTING APPLIED PROBLEMS USING MACHINE LEARNING
Keywords:
machine learning, academic performance forecasting, Random Forest, gradient boosting, fully connected neural networks, feature importance analysis, ROC analysisAbstract
The article explores data mining and predictive modeling methods for forecasting students' academic performance based on a complex of sociopedagogical factors. The authors perform a comparative analysis of the effectiveness of ensemble machine learning algorithms (Random Forest, XGBoost) and a deep fully connected neural network in solving regression and binary classification problems. Based on the trained models, a comprehensive framework for evaluating forecasting quality was established (including MSE metrics, F1-score, confusion matrices, and ROC-AUC), allowing for the effective identification of at-risk students to provide timely academic support.
References
1. R. J. Collie, J. D. Shapka and N. E. Perry, “School climate and social emotional learning: Predicting teacher stress, job satisfaction, and teaching efficacy”, Journal of Educational Psychology, vol. 104, no. 4, (2012), pp. 1189-1204.
2. C. Day, A. Kington, G. Stobart and P. Sammons, “The personal and professional selves of teachers: stable and unstable identities”, British Educational Research Journal, vol. 32, no. 4, (2006), pp. 601-616.
3. E. Deci, R. Vallerand, L. Pelletier and R. Ryan, “Motivation and education: The self-determination perspective”, Educational Psychologist, vol. 26, no. 3 &4, (1991), pp. 325-346.
4. J. DeCoster, “Overview of Factor Analysis”, Retrieved from http://www.stathelp.com/notes.html. (1998).
5. R. Delorenzo, W. Battino, R. Schreiber and B. Gaddy Carrio, “Delivering on the promise”, Bloomington, IN: Solution Tree, (2009).
6. B. Fuller and J. Izu, “Explaining school cohesion: What shapes the organizational beliefs of teachers”, American Journal of Education, vol. 94, (1986), pp. 501- 535.
7. C. H. Geer and J. Q. Morrison, “The impact of the collective efficacy of a school community on individual professional development outcomes”, MidWestern Educational Researcher, vol. 21, no. 4, (2008), pp. 32-38.
