Kaggle challenge: Santander
In this notebook we use the following concepts and methods:
- Undersampling and oversampling.
- Outlier detection.
- Anomaly detection with autoencoders.
- Gaussian Naive Bayes.
- Gradient boosted trees (XGB and LGBM), with grod-search and cross-validation.
- Bayesian hyperparameter optimization.
- Model blending for improved AUC scores.
If you have problems viewing the .ipynb notebook, go to: https://nbviewer.jupyter.org/github/peitsche/Santander/blob/master/kaggle_santander_git.ipynb