Skip to content
2nd place solution for the MeLi Challenge 2019
Jupyter Notebook
Branch: master
Clone or download
Pull request Compare This branch is even with tobiasveiga:master.
Type Name Latest commit message Commit time
Failed to load latest commit information.
MeLi_Ensembles added all files Oct 4, 2019
LICENSE added some readme info and fixed some file names Oct 4, 2019


2nd place solution for the MeLi Challenge 2019

This solution uses three models SGD (scikit-learn), multinomialNB (scikit-learn) and GRU (Keras) trained on both char level and word level, and in two different (but quite similar) datasets.

Instructions to generate submission

Required packages and softwares:

Python3 with the libraries Numpy, Pandas, Scikit-Learn, MatplotLib, Keras, NLTK. (No other library was used. No pre-trained model was used.)

Follow the steps:

  1. Unpack train, test and sample submission in the root folder with the respective names: train.csv, test.csv, sample_submission.csv

  2. Run all notebooks in MeLi_BaseGen/

  3. Run all notebooks in MeLi_scripts2/

  4. Run all notebooks in MeLi_scripts3/

  5. Run the notebook MeLi_Ensembles/MeLi_Ensemble_06.ipynb

  6. Run the notebook MeLi_Ensembles/MeLi_Ensemble_11.ipynb

  7. Run the notebook MeLi_Ensembles/MeLi_FinalEnsemble.ipynb

The final submission will be in the root folder with the name 'submission_MegaEnsemble02.csv' :)

PS: Some of these notebooks might require more than 64 GB of memory to run. Each notebook takes about 1 to 2 hours to run copletely in a 4 cores computer.

You can’t perform that action at this time.