Publicação
Predictive Modeling for Identifying Students at Risk of Academic Failure: A Comprehensive Approach
| Resumo: | This study develops a predictive model to forecast academic achievement among Portuguese students in grades 8 to 10, using a dataset sourced from DGEEC for the academic year 2021-2022, encompassing 262,874 students. The dataset includes variables such as student demographics, academic performance metrics, attendance records, and school characteristics. The thesis begins with an exploration of academic achievement (AA) and its influencing factors, reviewing previous research on AA drivers and the application of machine learning techniques. Methodology details encompass data preprocessing, feature engineering, and strategies for model enhancement, including SMOTE and feature selection. Key findings underscore the significance of student demographics, prior academic performance, and school characteristics in predicting academic outcomes. Noteworthy predictors identified include student age, gender distribution in schools, course grades, school size, and attendance patterns. The Random Forest (RF) model emerges as particularly effective, demonstrating high accuracy in identifying at-risk students, with targeted interventions proving successful for those in the highest risk decile, revealing a 55.53% fail rate. Strategies proposed for intervention encompass reducing class sizes to facilitate personalized learning, promoting peer study groups for collaborative learning, and incentivizing perfect attendance to enhance consistency and engagement. Leveraging online platforms for personalized learning and fostering parental involvement are additional recommendations aimed at supporting academic success across diverse student backgrounds. |
|---|---|
| Autores principais: | Castro, Inês Leitão Duarte de |
| Assunto: | Academic achievement Academic predictors Predictive Modelling Machine learning Educational data analysis SDG 4 - Quality education SDG 5 - Gender equality SDG 8 - Decent work and economic growth |
| Ano: | 2024 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | This study develops a predictive model to forecast academic achievement among Portuguese students in grades 8 to 10, using a dataset sourced from DGEEC for the academic year 2021-2022, encompassing 262,874 students. The dataset includes variables such as student demographics, academic performance metrics, attendance records, and school characteristics. The thesis begins with an exploration of academic achievement (AA) and its influencing factors, reviewing previous research on AA drivers and the application of machine learning techniques. Methodology details encompass data preprocessing, feature engineering, and strategies for model enhancement, including SMOTE and feature selection. Key findings underscore the significance of student demographics, prior academic performance, and school characteristics in predicting academic outcomes. Noteworthy predictors identified include student age, gender distribution in schools, course grades, school size, and attendance patterns. The Random Forest (RF) model emerges as particularly effective, demonstrating high accuracy in identifying at-risk students, with targeted interventions proving successful for those in the highest risk decile, revealing a 55.53% fail rate. Strategies proposed for intervention encompass reducing class sizes to facilitate personalized learning, promoting peer study groups for collaborative learning, and incentivizing perfect attendance to enhance consistency and engagement. Leveraging online platforms for personalized learning and fostering parental involvement are additional recommendations aimed at supporting academic success across diverse student backgrounds. |
|---|