Scholarlink

This study proposes a machine learning-based model, specifically using the Random Forest algorithm, to analyze the factors influencing the performance of engineering students in the Saber Pro test, administered by the Colombian Institute for Educational Evaluation (ICFES). The model is designed to integrate academic records, socioeconomic variables, and institutional data in order to identify and prioritize the key factors associated with student achievement. The proposed methodology includes data collection, preprocessing, feature selection, and hyperparameter optimization using GridSearchCV. Techniques such as categorical variable encoding, normalization, and cross-validation will be applied to ensure methodological rigour and model robustness. Rather than presenting experimental results, this proposal outlines the conceptual and technical foundation for developing a predictive tool that can inform evidence-based educational strategies. The ultimate aim is to support institutional efforts to enhance academic outcomes, reduce dropout rates, and provide a scalable analytical framework applicable to other higher education contexts.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identifying Key Drivers of Academic Performance in Colombian Engineering Students: A Machine Learning Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nathalia Orozco Morales,
  • Leonardo Valderrama Garcia,
  • Luis Rabelo,
  • Cristian Rincón-Guio

摘要

This study proposes a machine learning-based model, specifically using the Random Forest algorithm, to analyze the factors influencing the performance of engineering students in the Saber Pro test, administered by the Colombian Institute for Educational Evaluation (ICFES). The model is designed to integrate academic records, socioeconomic variables, and institutional data in order to identify and prioritize the key factors associated with student achievement. The proposed methodology includes data collection, preprocessing, feature selection, and hyperparameter optimization using GridSearchCV. Techniques such as categorical variable encoding, normalization, and cross-validation will be applied to ensure methodological rigour and model robustness. Rather than presenting experimental results, this proposal outlines the conceptual and technical foundation for developing a predictive tool that can inform evidence-based educational strategies. The ultimate aim is to support institutional efforts to enhance academic outcomes, reduce dropout rates, and provide a scalable analytical framework applicable to other higher education contexts.