How to Survive the Titanic

Machine learnings to avoid iceberg induced death.

Transcript

1. How to Survive the Titanic with Machine Learning
2. Titanic II …no, I’m not joking
3. How to Predict Survival
4. Pandas Make Life Better http://pandas.pydata.org/
5. What’s the data look like? 2200 Passengers 1500 Died 700 Survived age, sex, class, ticket price, survival
6. Human Models
7. Women and Children First
8. Sex Based Model if ‘male’ then Survived = false ~77% Accuracy
9. Adding More Variables sorry kids make way for the rich
10. Sex & Wealth Model if ‘male’ && ‘poor’ then Survived = double false ~79% Accuracy
11. Machine Learning Use half the data to develop model Use other half to test model accuracy Adjust the model according the results
12. Starting Decision Tree male dead 100%
13. Decision Trees dead alive dead alive dead alive male 3rd 2nd 1st 63% 37% 84% 16% 86% 14%
14. Random Forests
15. Machine Learning Results ~80% Accuracy
16. Conclusion Small data sets suck Simple models can be sufficient If you want to survive, don’t be poor or a man
17. Try It Yourself http://www.kaggle.com/c/titanic-gettingStarted http://scikit-learn.org/