Kaggle - Titanic Challenge

Machine Learning from Disaster

This project is inpired in the Titanic - Machine Learning from Disaster Kaggle Challenge to use data to predict whether a person that was in the Titanic survived or not to the disaster.

This is a well-known Kaggle challenge for begginners, and there is plenty of material available to learn and to improve your project. I worked in this project with the purpose of applying my new knowledge of Machine Learning and Neural Networks. With the material I found online I could learn new tools of the pandas, scipy and Scikit-Learn libraries.

The different programs try different approaches to the problem and achieve different accuracies.

Data format

The challenge data can be found here and it consists of the following features

• survival : 1 if the person survived, 0 otherwise. It is the variable to be predicted
• pclass: Descbribes the ticket class. It is a proxy for socio-economic status (SES)
  • 1st = Upper
  • 2nd = Middle
  • 3rd = Lower
• age: Age is fractional if less than 1. If the age is estimated, is it in the form of xx.5
• sibsp: The dataset defines the number of siblings/spouses aboard the Titanic family relations in this way:
  • Sibling = brother, sister, stepbrother, stepsister
  • Spouse = husband, wife (mistresses and fiancés were ignored)
• parch: The dataset defines the number of parents/children aboard the Titanic family relations in this way:
  • Parent = mother, father
  • Child = daughter, son, stepdaughter, stepson
  • Some children travelled only with a nanny, therefore parch=0 for them.
• ticket : Ticket number
• fare : Passenger fare
• cabin : Cabin number
• embarked : Port of Embarkation
  • C = Cherbourg, Q = Queenstown, S = Southampton

Models

• Titanic 1
  • Model: RandomForestClassifier from Sckit-learn
  • Features: Sex and Age to predict the model
  • Accuracy of 0.71291
• Titanic 4
  • Model: Testing over 8 different prediction models
  • More deep data analysis and Features engineering
  • Hyperparameters tuning with sklearn.model_selection.SRandomizedSearchCV
  • The best model was RandomForestClassifier
  • Accuracy of 0.78947

Acknowledgements

Special thanks to the article Predicting the Survival of Titanic Passengers by Niklas Donges for providing clear explanations and insights to the problem. Special thanks also to the YouTube channel Mario Filho - Data Science for the challenge tutorial and introduction to Kaggle.