In this repo you can find various attempts (tensorflow,catboost) to solve the Kaggle competition about fraud modeling, and also the data preparation needed before you can try the different models.
Here is the link to the Kaggle competition
The idea is to use machine learning to detect if a card transaction is fraudulent or not.
The best score I could get was from using Catboost models: this is a tree based algorithm developed by Yandex and open source Learn more by clicking here