Key points are not available for this paper at this time.
Abstract Predicting graduate employability has traditionally relied on surveys and statistical models, which struggle with high-dimensional, categorical, and imbalanced data. To address these gaps, machine learning (ML) offers improved predictive capabilities with detailed analytical insights. Graduate employability remains a pressing concern, particularly in sub-Saharan Africa (SSA), where conflicts, epidemics, and fragile economies amplify unemployment challenges. This study investigates IT graduates employability in a low-income sub-Saharan African (SSA) country and proposes a tree-based machine learning classifier for predicting graduates employment. The four tree-based machine learning algorithms are compared using ten-fold and five-fold cross-validation methods, as well as multi-feature attribute evaluators, information gain, and the chi-square test. The multi-feature evaluation results indicate that factors such as marital status, IT competence, IT specialisation, working experience, and age influence the employability of IT graduates in the SSA country. The random forest, decision tree, AdaBoost, and the random tree algorithms were evaluated using machine learning metrics including, accuracy, F-measure, and confusion matrix to ascertain dominant performance. The random forest emerged as the most effective algorithm, achieving accuracy, F-measure, and ROC-AUC values of 88.11%, 0.964, and 0.945, respectively. The second most effective classifier is the random tree algorithm, which attained an accuracy of 86.76%. This study is significant as it supports universities in aligning curricula with employability skills, assists employers in refining recruitment strategies, and provides policymakers with evidence-based insights to address graduate unemployment. The relevance of these contributions is particularly pronounced in low income and conflict affected countries within sub-Saharan Africa.
Dake et al. (Wed,) studied this question.