This repository represents several projects completed in IE HST's MS in Business Analytics and Big Data's Financial Analytics course.
This was a kaggle style project whereby the main objective was to use financial machine learning best practices and innovative techniques to move up the leaderboard of the Santander Customer Transaction Prediction.
Model and Results
- Used a Light GBM and Hyper Parameter Fine Tuning.
- Used the Frequency Distribution for each Column as Feature / Data Augmentation.
- Created a separate model per Variable and then found coefficients with Logit to ensemble the predictions.
The different approachs and rationale are documented in Finance_GroupCompetition_Santander.ipynb.
Our final position in the Public Leader Board would be: 59 with an AUC score: 0.92244
This project was completed with Guillermo Chacon, Diego Cuartas, Esteban Delgado, Aayush Kejriwal, Denis Mochalov, Mariana Narvaez and Valentina Premoli.