Design of a Predictive Model Based on Neuromathematics and Machine Learning to Anticipate Academic Performance

Abstract

The aim of this study was to design a predictive model based on Neuromathematics and Machine Learning to anticipate academic performance in a geometric task. Using the facial recognition software FaceReader, the micro-expressions of 426 students were captured in real time during the construction of a cone in the Cabri Express software. The model achieved 91% accuracy by integrating emotional data and participants' sex as predictive variables. The results indicated that emotions such as happiness and surprise were positively correlated with successful performance, while anger and neutrality were associated with unsuccessful performance. Additionally, significant differences between men and women were identified, highlighting the importance of including sex in the analysis. Neuromathematics allowed the application of knowledge about brain mechanisms in mathematical learning, also guiding pedagogical and didactic processes. Although the study focused on a specific geometric task, its implications are broad, as this approach can be applied to other educational contexts. Among the limitations, the lack of generalization to other mathematical areas is noted, but it is proposed to explore its applicability in future studies in disciplines such as algebra or calculus. This predictive model offers a valuable tool for personalizing teaching, adjusting pedagogical interventions based on emotions and sex, thus improving academic performance and student well-being.