Logistic regression is a reliable model for predicting licensure exam performance of future math teachers

23 Dec 2025

UP Diliman

In the Philippines, passing the Licensure Examination for Teachers (LET) is a crucial step in becoming a professional teacher, but the passing rate has been declining—from 31.45% in 2010 to just 27.28% in 2018. To address this issue, a study was conducted to predict the performance of future math teachers in the LET using machine learning. The researchers tested three different algorithms—Gradient Boosted Trees, Logistic Regression, and Naïve Bayes—using data from four universities with 769 data points. After evaluating the models based on accuracy, precision, and other performance metrics, all three models showed good results, but Logistic Regression was found to be the most reliable. This model performed best when applied to evaluation data, making it the most suitable for predicting LET outcomes.

The significance of the study lies in its potential to substantially improve teacher preparation and the overall quality of education in the Philippines. By utilizing machine learning to predict the performance of future math teachers on the Licensure Examination for Teachers (LET), the research offers an innovative, data-driven approach to understanding factors that contribute to success or failure in the exam. The predictive models, particularly Logistic Regression, can help higher education institutions (HEIs) identify students who may need additional support, enabling more targeted interventions and personalized learning strategies. This could result in better preparation for the LET, leading to higher pass rates.

Additionally, the study provides valuable insights that can guide the refinement of curricula, teaching methodologies, and resource allocation within teacher education programs. The findings also provide insights into the current state of prospective mathematics teachers and offer recommendations for improvement. The ability to predict outcomes allows HEIs to be more proactive in their approach, focusing on areas of improvement before students take the exam. Furthermore, the findings can inform policy development in the education sector, leading to improved teacher quality and more effective licensure processes. Ultimately, the study underscores the importance of leveraging machine learning to drive positive change in teacher education, ensuring that prospective teachers are better equipped for the challenges they will face in the classroom.

Authors: Arturo J. Patungan (College of Education, University of the Philippines Diliman | Department of Mathematics and Physics, College of Science, University of Santo Tomas) and Ma. Nympha B. Joaquin (College of Education, University of the Philippines Diliman)

Read the full paper: https://doi.org/10.1063/5.0230585

Logistic regression is a reliable model for predicting licensure exam performance of future math teachers

In the Philippines, passing the Licensure Examination for Teachers (LET) is a crucial step in becoming a professional teacher, but the passing rate has been declining—from 31.45% in 2010 to just 27.28% in 2018. To address this issue, a study was conducted to predict the performance of future math teachers in the LET using machine learning. The researchers tested three different algorithms—Gradient Boosted Trees, Logistic Regression, and Naïve Bayes—using data from four universities with 769 data points. After evaluating the models based on accuracy, precision, and other performance metrics, all three models showed good results, but Logistic Regression was found to be the most reliable. This model performed best when applied to evaluation data, making it the most suitable for predicting LET outcomes.

The significance of the study lies in its potential to substantially improve teacher preparation and the overall quality of education in the Philippines. By utilizing machine learning to predict the performance of future math teachers on the Licensure Examination for Teachers (LET), the research offers an innovative, data-driven approach to understanding factors that contribute to success or failure in the exam. The predictive models, particularly Logistic Regression, can help higher education institutions (HEIs) identify students who may need additional support, enabling more targeted interventions and personalized learning strategies. This could result in better preparation for the LET, leading to higher pass rates.

Additionally, the study provides valuable insights that can guide the refinement of curricula, teaching methodologies, and resource allocation within teacher education programs. The findings also provide insights into the current state of prospective mathematics teachers and offer recommendations for improvement. The ability to predict outcomes allows HEIs to be more proactive in their approach, focusing on areas of improvement before students take the exam. Furthermore, the findings can inform policy development in the education sector, leading to improved teacher quality and more effective licensure processes. Ultimately, the study underscores the importance of leveraging machine learning to drive positive change in teacher education, ensuring that prospective teachers are better equipped for the challenges they will face in the classroom.

Authors: Arturo J. Patungan (College of Education, University of the Philippines Diliman | Department of Mathematics and Physics, College of Science, University of Santo Tomas) and Ma. Nympha B. Joaquin (College of Education, University of the Philippines Diliman)

Read the full paper: https://doi.org/10.1063/5.0230585