Notebook for the Titanic kaggle competition. The goal of the competition is to predict passengers life status after the sinking. We are provided we numerous of variables, continous and categorical.
We start by cleaning the data, and we building 2 models. One based on XGBoost and an another using deep learning in Keras. We put a great emphasis on the models interpretation, using partial dependency plot, and the models evaluations using important indices (F1 score, precision/recall, ROC curve,...).
Both models are in the top 10% of the Kaggle competition.