This thesis explores the use of Machine Learning (ML) techniques to model and predict transportation mode choice behavior, a critical component of urban mobility planning. Traditional mode choice modeling relies on Random Utility Maximization (RUM) theory, with models such as Multinomial Logit model and Mixed Logit model (MXL). While these offer interpretability, they often struggle with complex feature relationships and heterogeneity in large datasets. ML methods, by contrast, offer greater predictive power and flexibility, albeit with interpretability challenges. This study evaluates the performance of various ML models like Gradient Boosting and Random Forest Decision Tree on two datasets, with a primary focus on a detailed case study using Swiss travel data. The models are assessed under various configurations, including feature limitation, latent variable extraction, and SMOTE-based resampling. Comparative results demonstrate that ML models consistently outperform traditional Logistic Regression models in terms of F1 and Balanced Accuracy metrics. Additionally, tools like SHapely Additive exPlanation (SHAP) are employed to enhance the interpretability of ML outcomes. The findings highlight the potential of ML to improve Mode Choice modeling, particularly when combined with theory-informed structures and advanced data balancing techniques.