Academic Risk Prediction: An Artificial Intelligence-Based Approach Using Psychoeducational Variables

This study presents an academic information management model based on Artificial Intelligence (AI) for early predicting school failure risk. It integrates classical statistical techniques, such as multiple regression and correlation analysis, with advanced machine learning methods, including K-means clustering and Principal Component Analysis (PCA). This methodological combination enables a comprehensive analysis of data related to cognitive skills, executive functions, physical and mental health, emotional well-being, and sociodemographic factors. The model was applied to a sample of 190 students aged 8 to 12 from vulnerable communities in Colombia, with data collected at three key points: mid-school year, end of the cycle, and the beginning of the following academic year. The analysis achieved a predictive accuracy of 85%, highlighting the importance of mental health indicators, especially depression and anxiety, in predicting academic performance and reading comprehension. Significant interactions between emotional and cognitive variables were found, underscoring the need for integrated approaches when designing effective educational interventions. This approach allows for more targeted preventive actions and supports ongoing evaluation of the model’s stability over time. Incorporating more sophisticated AI techniques, such as deep neural networks and boosting models, is proposed as a future direction to enhance the model’s predictive capacity and broaden its applicability across diverse educational contexts.