Скачать или смотреть How I Built a Fraud Detection Model with AMEX Data Using XGBoost

In this video, I walk you through my project on fraud detection using the AMEX Kaggle dataset with XGBoost. I demonstrate how I reduced the dataset size from 50GB to 3.5GB using Parquet and RAPIDS, performed feature selection with ANOVA, correlation, and linear regression, and compared models using the top 20% of features versus all features.

You'll see how optimizing the feature set improved model speed and reduced the risk of overfitting, achieving an accuracy of 0.84 and a recall of 0.90. Check out my GitHub for the complete code and connect with me on LinkedIn for more projects. Don't forget to visit my YouTube channel for other videos like building a data warehouse in SQL with Python!

Links:

GitHub: https://github.com/simon-gobin/Amex-F...
LinkedIn: www.linkedin.com/in/simongobin
Website: simongobin.wordpress.com

Thanks for watching!