This is my work on the Kaggle's competition to detect credit card fraud transaction.
The detail of the competition can be found on [https://www.kaggle.com/c/ieee-fraud-detection/overview](Kaggle's website)
This is the work I submitted for the competition. First I did an exploratory data analysis to understand the data, and then I build a model based on the train dataset, using lightGBM to predict the fraud on the test dataset provided.
I entered the competition late and didn't have much time, so there are few features engineering, no blending, and the final model submitted scored 0.941296 in LB and 0.915805, which place it 2603 on 6381 teams (first 41%)
This is a rework in progress with no time pressure, learning from the mistakes I did, and using the insights given by the other competitors.