Article

The Impact of Educational Profiles on Salary Levels among Employed Nigerian Graduates: A Machine Learning Analysis with Random Forest and Gradient Boosting

---

Abstract

This study investigates the relationship between educational profiles and salary levels among employed Nigerian graduates using interpretable machine-learning models. A dataset of 3,000 respondents from the Nigerian Graduate Survey was analyzed through Random Forest and Light Gradient Boosting Machine (LightGBM) classifiers. Ten demographic and educational attributes—including age, gender, region, field of study, GPA, university type, and postgraduate qualification—were used to predict salary level categories (“Low-paid” and “Well-paid”). Data preprocessing involved one-hot encoding for categorical variables and stratified training–testing splits to ensure balanced evaluation. Results indicated that LightGBM slightly outperformed Random Forest, achieving an accuracy of 0.571, compared to 0.567 for Random Forest. Both models exhibited strong recall for the “Low-paid” category but struggled with precision on “Well-paid” graduates, reflecting the dataset’s class imbalance and the influence of external labor-market factors beyond education. Feature importance analysis identified field of study, university type, and postgraduate degree as dominant predictors of salary outcome. These findings suggest that while education remains a critical driver of earning potential, its impact is mediated by broader socio-economic variables not fully captured within the dataset. The study highlights the potential of machine learning to generate actionable insights for aligning higher-education curricula with labor-market demands, thereby enhancing graduate employability and economic equity.

Keywords: Educational attainment, Salary prediction, Nigerian graduates, Machine learning, Gradient boosting

How to Cite:

Guballo, J. O., (2025) “The Impact of Educational Profiles on Salary Levels among Employed Nigerian Graduates: A Machine Learning Analysis with Random Forest and Gradient Boosting”, Artificial Intelligence in Learning 1(4), 287-300. doi: https://doi.org/10.63913/ail.v1i4.42