Our project tries to predict the probability that an online transaction is fraudulent. We focus on the XGBoost method.
This repository contains our final work for the Advanced Machine Learning course at ENSAE, Institut Polytechnique de Paris.
In this repository, you will find the following files:
- A notebook (.ipynb), explaining in more depth our modeling approaches. This notebook contains the main results.
- A database (.zip) in the input folder that was obtained from the IEEE Computational Intelligence Society (IEEE-CIS).
If you want to run the notebook from your computer, you will need to clone this repository.
- The IEEE-CIS Fraud Detection database which we used in this project is available for access at the link: https://www.mavenanalytics.io/data-playground](https://www.kaggle.com/competitions/ieee-fraud-detection/discussion/100662.
We would also like to thank Professor Austin J. Stromme for the support and knowledge they gave us.
In the contemporary landscape of online transactions, fraud detection plays a pivotal role in ensuring financial security. the Institute of Electrical and Electronics Engineers' Computational Intelligence Society (IEEE-CIS), in collaboration with Vesta Corporation, organized a Kaggle competition dedicated to improving fraud prevention systems. This competition invited participants to develop and benchmark machine learning models using a substantial dataset derived from real-world e-commerce transactions provided by Vesta. The dataset encompasses diverse features, ranging from device types to product characteristics.
The objective of a well-performing model is to significantly enhance the accuracy of fraudulent transaction alerts, benefiting millions of consumers globally. The successful implementation of such models contributes to the reduction of fraud losses for businesses, ultimately leading to increased revenue. This collaborative effort is geared towards making fraud prevention more effective, providing a seamless and secure experience for users engaging in online transactions.