ak9969 / kaggle-titanic-octave

kaggle-titanic-octave

Kaggle Titanic project - http://www.kaggle.com/c/titanic-gettingStarted This repo contains files used to generate a solution to the Kaggle Titanic getting started project. The solution files are written in Octave. The core of the solutions are borrowed from the Cousera "Intro to Machine Learning" class taught by Andrew Ng. "test_original.csv" and "train_original.csv" are the original files downloaded from the kaggle website. The "_modified.csv" versions of these files are the ones I modified to contained no hearders and only the paramter values of interest to the algorithm.

In the "train_modified.csv" the columns represented are survived, pclass, sex, age, sibsp, parch, cabin. I adjusted these columns to be numeric values (in cases where they weren't) and simplified some of the range of values to make the algorithm work better. For "sex" the values are 0=male, 1=female. For "age" I simplified the values to indicate only adult or child (age >= 18 => 1, age < 18 => 0). If the age date was missing, I assumed the passenger was an adult and assign a 1. The "cabin" values ranged from "A" to "F" so I merely substituted 1-6, respectively, and 0 if the cabin data was unknown.