dkozl50 / pluh_sms Goto Github PK

This repository contains the code that was used for dataset analysis and the models optimizing uplift of test and control groups. The dataset was provided to us by the hackathon organizers at Lenta and Microsoft. The structure of the repository is the following:

1. experiments - The folder with jupyter notebooks used for dataset analysis and model tuning.
2. features contains functions for feature engineering.
3. scripts is the folder with some utility functions. Models are located in pipeline. The final model is BlendedModel that uses blending of several independent algorithms.
4. submissions - the name speaks for itself.
5. site - The source code of our dashboard.