Previsão da procura de estudantes utilizando modelos de regressão: um estudo comparativo da unidade académica de engenharia, indústria e construção da Universidade Católica de Cuenca
Year: 2025
Type: article
Abstract: Introduction: This study develops a methodology for forecasting student demand at the Academic Unit of Engineering, Industry, and Construction of the Catholic University of Cuenca, addressing the challenge of short time series. Materials and methods: Daily enrollment data (22 observations/program) were used. To overcome data scarcity, data augmentation was applied by combining SMOTE interpolation with bootstrapping, expanding the sample. Four regression models were evaluated on this augmented data: linear, polynomial (2nd and 3rd degree), and logistic. The selection of the optimal model for each of the five programs was performed using a multi-criteria framework that integrates the Neutrosophical Analytic Hierarchy Process (AHP)—to weight criteria such as R², RMSE, MAPE, and MAE under uncertainty—and the TOPSIS technique for the final ranking. Results: The data augmentation allowed the series to be expanded to approximately 55 observations. The TOPSIS analysis revealed that there was no single optimal model: the third-degree polynomial model was selected for Civil and Industrial Engineering; the second-degree model for Architecture and Electrical Engineering; and the linear model for Design. Discussion: The proposed methodology proves superior to the use of deep learning with limited data, avoiding overfitting. The combination of data augmentation and neutrosophic multi-criteria selection provides a robust, transparent, and adaptable framework that manages uncertainty and justifies model selection based on the specific dynamics of each program. Conclusions: This study validates that data augmentation and model selection using Neutrosophic AHP-TOPSIS constitute an effective and pedagogical tool for demand forecasting in data-constrained contexts, offering university management a tool for more precise and evidence-based academic and resource planning.
Source: Maestro y sociedad
Language: Spanish
Field: Decision Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: diamond
OpenAlex ID: W7220199373
