Using historical data from the Oracle EAM platform, the research follows a structured experimental approach involving data preprocessing, feature engineering, and iterative testing of multiple algorithms. The primary goal is to assess the predictive performance of classical machine learning models Logistic Regression, SVC, Linear SVC, Decision Tree Classifier, and Random Forest Classifier alongside deep learning models of varying complexities (small, medium, and large). To address the imbalance in active vs. cancelled vehicles, the F1 score was used as the main evaluation metric. The Random Forest Classifier achieved one of the best results with an F1 score of 0.882, showing robust performance across multiple iterations. The Decision Tree Classifier also performed strongly, highlighting the potential of ensemble learning for this use case. Among the deep learning models, the small model outperformed others, achieving an F1 score of 0.880 on the test set, validating its suitability for structured, tabular datasets with limited complexity.