Design and development of a machine learning based personalized career recommendation platform (WEB-MÖS) for computer engineering graduates

In this study, a web-based intelligent system has been designed and developed to provide personalized career recommendations for computer engineering students based on their academic profiles and technological interests. The project comparatively analyzes two different machine learning modeling approaches. In the first stage (Model-1), it was determined that models trained with a purely rule-based synthetic dataset had limited generalization performance when tested on real student data from 50 individuals, achieving a maximum accuracy of 52% (Logistic Regression). To overcome this generalization problem, a hybrid data augmentation strategy was implemented in the second stage (Model-2), which utilized the real dataset of 50 individuals and enriched it using the class-specific CTGAN (Conditional Tabular Generative Adversarial Network) technique. Models such as Random Forest, Gradient Boosting, and Support Vector Machine (SVM), trained with the synthetic weighted hybrid dataset generated by this method, achieved an accuracy rate of over 97% on the test set. This result confirms that data augmentation with CTGAN, starting from a small amount of real data, significantly enhances the model's learning capacity and predictive power. The web application (WEB-MÖS), developed with the Python Flask framework for deployment, not only serves real-time predictions but also features a dynamic structure that allows for the periodic retraining of the model by collecting user feedback through Google Sheets API integration. This holistic approach presents an effective career recommendation system prototype that provides data-driven guidance for students in the field of computer engineering education.