Titanic - Predict Survival on the Titanic

Highest Accuracy: 0.794

​

Models:

1. Random Forests Classifier and Gradient Boosting Classifier as 1st Level Models

2. K Nearest Neighbours Classifier as 2nd Level Models

​

Takeaways:

1. Smart Feature Engineering of combining existing features or dropping features often beats having an over-complicated model

​

2. Always helps to use logical or contextual knowledge to create new features (etc. grouping passengers by surname as families would tend to stick together)

​

3. Always standardize your features and use hyperparameter tuning for models