Drop of Labels unnecessary

[AshwinB-hat]

Hey,
The reason you have such a low accuracy even when you have the proper feature-train ratio is because many of the features that you are dropping has direct correlation with the output.
Please have a look at this kernel on feature engineering on the titanic.
https://www.kaggle.com/harryem/feature-engineering-on-the-titanic-for-0-81339

[pranay414]

Thank you :), I'll check it out and incorporate it in my code.