As of October 2018, I managed to get to top 8% in the Titanic Kaggle competition (751 out of 9925 participants).

In the solution I use a RandomForestClassifier as base estimator for GridSearchCV to tune different hyperparameters. I also, included some new engineered features.