This competition aims to predict survivals on the Titanic.
I have written a R function to parse the training dataset. In fact, I would launch the learning algorithm on specific data composition : eg- RandomForest on Men || > 30years || with no familly member || from class n° 1..
Nevertheless, to go too farther in detail may imply overfitting issues. We should make the optimal tradeoff between details and overfitting.
Finnaly, within my function we can choose the learning algorithm and aggregate prediction for each sub datasets.
I was ranked 200th / 1200.
Feel free to contact me for any question/suggestions.
Enter text in Markdown. Use the toolbar above, or click the ? button for formatting help.