Student job readiness is a crucial issue that requires serious attention in the university environment. This problem is crucial because universities need to identify existing students whose job readiness is low to prepare them to enter the workforce. The manual detection process has significant limitations, such as the need for psychologists, which requires substantial time, effort, and money. Therefore, implementing machine learning can address these challenges more effectively. This study aims to analyze and compare the performance of two algorithms, Multi-Layer Perceptron (MLP) and Extreme Gradient Boosting (XGBoost), in predicting student job readiness using a dataset of XYZ University graduates, which includes student data and post-graduation job readiness surveys. Additionally, this study seeks to evaluate the effectiveness of these models compared to previous research findings, specifically those obtained using the Multinomial Logistic Regression model. The results showed that the XGBoost model consistently performed better than the MLP model at various data-sharing ratios, with XGBoost achieving an accuracy of 89% compared to MLP 82%. Furthermore, XGBoost precision, recall, and F1-Score are higher than those of MLP, indicating XGBoost superior ability to generalize models. Enhancing performance is achieved using SMOTEENN technique, which aids in balancing the data and boosting the model's effectiveness. The comparison with previous research, where the best Multinomial Logistic Regression model achieved an accuracy of 53.9%, highlights the significant improvement in accuracy with XGBoost and underscores the effectiveness of modern machine learning techniques in classifying students’ job readiness based on when they got a job.
Keywords—multi-layer perceptron (MLP), extreme gradient boosting (XGBoost), hyperparameter, job readiness, smoteenn