Predicted train punctuality out of a data science approach with classification primary focused on western Sweden. Extracted real data from Trafikverket's API that included properties such as trains, train stations, deviation codes, deviation information, train delay in minutes and train punctuality. Transformed the data with PySpark in Databricks, while also creating visualizations in Databricks to explore the data and come up with findings. Loaded the data into Jupyter Notebook and built 4 machine learning models with Spark ML, and compared them, evaluated the models, hypertuned the models and chose the one that performed best and the most consistent model for train punctuality predictions.