@KuntaKinte Data leaks are 'normal' but not so common in data and its upto the participants to detect this otherwise private leaderboard score will be worse than the public. if you have done Kaggle competitions then you must know this.
@steph_en_DataHack4Fi How do you get data leaks from data you don't have yet - test data isn't provided to participants. I think I'm missing your point, please explain.
@Ivor that type of leakage is a target leakage meaning that the predictors include data that won't be available at the time of prediction. To handle such a leakage, any variable that's created after the target value is realized should be excluded in training.
Hi everyone. Sorry for the delay on weighing in! @yohza37's makes a very good point here. We want to reassure you there is no data leakage and no "trick". Keep up the good work on this challenge!
Given that there only under 100 fraud cases the possibility arent that slim
Let's see how Zindi responds, otherwise its game over!
@KuntaKinte Data leaks are 'normal' but not so common in data and its upto the participants to detect this otherwise private leaderboard score will be worse than the public. if you have done Kaggle competitions then you must know this.
@steph_en_DataHack4Fi How do you get data leaks from data you don't have yet - test data isn't provided to participants. I think I'm missing your point, please explain.
@Ivor that type of leakage is a target leakage meaning that the predictors include data that won't be available at the time of prediction. To handle such a leakage, any variable that's created after the target value is realized should be excluded in training.
Thanks for explaining it. Much appreciated.
Hi everyone. Sorry for the delay on weighing in! @yohza37's makes a very good point here. We want to reassure you there is no data leakage and no "trick". Keep up the good work on this challenge!
Thanks Zindi