Machine learning models and techniques applied to CTGAN-generated data

This work describes investigations and results obtained using different gradient boosting algorithms, feature engineering techniques, and other machine learning techniques applied to a collection of Deep Learning based Synthetic Data Generators for single table data. Specifically, experiments shown were performed on tabular data provided by the Kaggle's 30 Days of ML Competition. Our method of Ordinal Encoding, boosting, hyperparameter optimization, ensembling, and stacking led us to reach 6th place in the competition out of 7573 teams worldwide among about 113,500 initial participants attending the connected course of Machine Learning.