Predictive Modelling of Teachers’ Workload, Classroom Management Practices and Students’ Factors on SS2 Students’ Academic Performance in Chemistry in Southwest Nigeria
No Thumbnail Available
Date
2026-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Lead City University, Ibadan
Abstract
This study investigated predictive modelling of Teachers’ Workload, Classroom Management Practices and Students’ Factors on SSS2 Students’ academic performance in Chemistry in Southwest Nigeria. Since most of the variables in behavioural science are categorical variables, there is need to use statistical tools that would correctly categorize data into different groups and predict academic performance from data from which reliable deductions could be drawn. Also, there is dearth of studies on predictive capabilities of AI algorithms in the prediction of students’ academic performance of students especially in African contexts. Consequently, the study leveraged machine learning algorithms to investigate the features that could contribute mostly to the performance of machine learning models in the prediction academic performance of students in Chemistry. Four research questions guided the study, and three Theories, Job Demand-Control Theory, Behaviourist Theory and Social Learning Theory provided the
needed insights into the study. The study adopted cross sectional research design. The population of the study comprised all the senior secondary school 2 students in the Southwest Nigeria. Multistage clustering sampling technique was used for the selection of participants. The process of data collection involved clustering of six states into three states in the southwest Nigeria. Each of the clustered states was stratified into three senatorial districts, and all the chemistry teachers in the sampled schools were selected for the study. Four research instruments were developed for data collection which include Teacher Workload Rating Scale (TWRS), Teacher Classroom Management Skills Questionnaire (TCMSQ), Students Factors
Questionnaires (SFQ), and Chemistry Achievement Test (CAT). The reliability coefficients of the instruments were established by using Cronbach Alpha with reliability coefficients in the range of (0.6-0.75). Data collected was analysed by using machine learning algorithms which include Gradient Boosting Model, Confusion Matrix and SHAP analysis. Results revealed that XGBoost achieved the highest overall test accuracy (0.6622) and a cross-validated F1-score (0.6313± 0.01422). However, Gradient Bosting exhibited strength in multi-class prediction F1-score (0.6524). It also achieved the highest Test ROC AUC (0.7714) which indicated a high performance accuracy in the classification of students on achievement basis. The most influential feature that contributed to the performance of the model in the prediction of students’ academic performance in chemistry was Teacher Non-Academic Workload (0.087354). Also, the study revealed that the model correctly predicted the academic performance of 160 students
accurately. SHAP analysis (Shapley Additive exPlanations) revealed that Teacher Non Academic Workload was the most impactful feature in the prediction of student chemistryperformance. The study concluded that a teacher’s daily workload is a far more powerful driver of students’ success. The study recommended among others that urgent intervention is needed to address the issue of teacher’s workload in secondary schools.
Keywords: Academic Performance, Teachers’ Workload, Management Skills, Students’ Factors
Word Count: 439
Description
Keywords
Academic Performance, Teachers’ Workload, Management Skills, Students’ Factors
Citation
Kate Turabian