Predictive Modeling of Graduate Vocational Mobility Using Multivariate Attributes

In an era marked by rapid labor market transformations and increasing demands for employability, the ability to accurately predict the vocational mobility of graduates holds significant value for educational institutions, policymakers, and career counselors. This study presents a data-driven framework for forecasting graduate vocational mobility using multivariate attributes derived from academic, demographic, experiential, and psychosocial domains. Four machine learning algorithms—Logistic Regression, Random Forest, Support Vector Machine (SVM), and Artificial Neural Network (ANN)—were implemented and evaluated using a labeled dataset of 10,000 graduate profiles. Comprehensive preprocessing, including data cleaning, feature engineering, and balancing techniques, was applied to ensure model readiness. Performance evaluation was conducted using accuracy, precision, recall, F1-score, and ROC-AUC, with cross-validation employed to assess robustness. The ANN model outperformed others across all metrics, demonstrating superior accuracy (91.3%), robustness (±0.009 in F1-score), and fairness across gender, region, and socioeconomic status subgroups. Feature-importance analysis revealed that attributes such as internship participation, GPA, communication proficiency, and adaptability were the most influential predictors of career mobility. The study also conducted subgroup fairness assessments to ensure ethical deployment and minimize bias. Overall, the proposed predictive framework provides a scalable and interpretable tool for guiding graduate career pathways. It enables institutions to make data-informed interventions and promotes equitable decision-making in workforce planning and graduate support programs.