In this repository you could find solution of the 2019 Data Science Bowl competition on Kaggle platform. We took 39-th place out of 3497 participants (Top 2%, silver medal).

Big respect to my teammates - Tatiana Glazkova and Andrew Ponomarev

It was really hard to avoid shake-up and get stable model. Look at right picture (thanks to James Trotman for pic.). It clearly shows the problem..

How to reproduce solution:

1. Download input data from https://www.kaggle.com/c/data-science-bowl-2019/data
2. Create directory with name "input" in root directory
3. Unzip all files to "input" directory
4. Open ./scripts/final_solution.ipynb
5. Run this notebook
6. Wait...
7. Answers will be stored in "submission.csv"