Titanic Survival Prediction

An algorithm that determines whether or not you're likely to have survived the Titanic (to an accuracy of 77%), based on parameters such as age, fare, number of dependents on board, and sex.

Try it interactively here.

Structure

• app.py is the main file. It is a Flask server that accepts a GET request containing the necessary parameters, and returns the appropriate response.

• train.py is the python program that trains the DecisionTreeClassifier on the data in train.csv. The DecisionTreeClassifier model object is then pickled into the model.mdl file. A StandardScaler is also trained on the data in train.csv and is then pickled into scaler.mdl to be used to scale the parameters input in app.py.

Use

To use this algorithm, there are two options:

• Interactive

  I've made a webpage that allows you to input the required parameters in a form. The webpage then outputs the result on submission of the form. Try it here.

• GET request

  The flask server is hosted at https://darrendube.pythonanywhere.com/.

  The required parameters in the GET request are:

  • age - int
  • sex - string ('female' or 'male')
  • fare - int (either 13, 20, or 83 - approximately $1300, $2000, or $8300 in today's US$)
  • parents - int (number of parents on board)
  • siblings - int (number of siblings on board)
  • spouse - int (0 = spouse is not on board. 1 = spouse is on board)
  • children - int (number of children on board)
  • title - `string ('Mr', 'Mrs', 'Miss', or 'Master')

  An example of a request would be https://darrendube.pythonanywhere.com/?age=36&sex=female&fare=83&parents=0&siblings=0&spouse=0&children=0&title=Mrs.

  The server either returns:

  • [0] - likely would not have survived, or
  • [1] - likely would have survived