Kaggle 'Search Results Relevance' 2nd place solution

Mikhail Trofimov, Stanislav Semenov, Dmitry Altukhov

Gets score 0.71881 on private leaderboard

How to reproduce submission

Don't forget to check paths in ./cfg.py! By default, you should place raw data in ./raw/, create ./processed/ for temporal files and ./submit/ for submission file.

After this, run

python preprocessing_mikhail.py
python preprocessing_dmitry.py
python preprocessing_stanislav.py
python fit_models_dmitry.py
python fit_model1_mikhail.py
python fit_model2_mikhail.py
python blend.py

Dependencies

• python 2.7.6
• numpy 1.10
• pandas 0.16.0
• scikit-learn 0.16.1
• scipy 0.15.1
• nltk 3.0.3
• BeautifulSoup 4.3.2
• tsne 0.1.1 (https://github.com/danielfrg/tsne)
• gensim 0.11.1-1
• backports.lzma 0.0.3

Hardware

This code was developed on a machine with 32 cores and 60 gb ram (amazon cc2.8xlarge instance), however it is possible to build a solution even with the average laptop.