A New Methodology for Predicting Teachers' Job Satisfaction Using Machine Learning Approaches - Bangladesh Perspective

Abstract

Shortage of qualified teachers and higher turnover rates have far-reaching negative effects on students' achievement and overall educational quality. Since satisfied teachers are less likely to quit their jobs and are more likely to deliver high-quality teaching, scholars around the world are trying to better understand the factors that predict teachers' job satisfaction. However, relatively little research employing machine learning (ML) techniques for detecting teachers' job satisfaction has been reported, particularly in developing countries like Bangladesh. Despite having a far-reaching effect on teaching quality which eventually impacts on education, there is a little work done in this regard in Bangladesh, specially in primary and secondary education sector. To bridge this gap, the present study has introduced a machine learning-based expert system framework for predicting teachers' job satisfaction. Data was gathered from 297 teachers of 56 different primary and secondary schools in Bangladesh. Missing values were imputed using multiple imputation techniques and feature selection techniques were adopted to find out the key affecting factors. To find the optimal model, ten supervised machine learning classifiers (Decision Trees, Support Vector Machines, Logistic Regression, K-nearest Neighbors, Random Forests, Gaussian Naive Bayes, Gradient Boosting, AdaBoost, Multi-layer Perceptron, and Quadratic Discriminant Analysis) were trained and tested using five-fold cross-validation. The performance of the classifiers was evaluated using multiple evaluation metrics including confusion matrix, accuracy, precision, recall, F1 score, and roc curve analysis. Findings revealed that Random Forest outperforms the other classifiers with 94% accuracy. The proposed framework is expected to assist the devolved authorities in implementing synchronized policies to improve teachers’ job satisfaction and curve analysis. Findings revealed that Random Forest outperforms the other classifiers with 94% accuracy. The proposed framework is expected to assist the devolved authorities in implementing synchronized policies to improve teachers’ job satisfaction and curve turnover rates