https://github.com/rkcosmos/santander-product-recommendation

https://github.com/rkcosmos/santander-product-recommendation

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/rkcosmos/santander-product-recommendation
• Owner: rkcosmos
• Created: 2017-01-06T02:53:44.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2017-05-24T04:29:28.000Z (over 7 years ago)
• Last Synced: 2024-09-17T16:50:04.432Z (4 months ago)
• Language: Jupyter Notebook
• Size: 104 KB
• Stars: 7
• Watchers: 2
• Forks: 10
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Santander Product Recommendation

This is my first Kaggle competition. The best model is in the first top 7%. You can read the high-level explanation in Thai here (http://kittinaradorn.com/first_kaggle_competition/).

**Instruction:**

1. Download dataset from https://www.kaggle.com/c/santander-product-recommendation/data, unzip and put them into input folder.

2. Create preprocessed data by running all cells in the following notebooks

  - MakeJuneExtraData.ipynb

  - MakeJuneExtraDataMulticlass.ipynb

  - MakeDataMulticlass2.ipynb

  - MakeTestDatawithpast3.ipynb

3. Choose model to create prediction from the following notebooks:

  - Baseline model: run MostProbableProductRecent2.ipynb

  - Basic logistic regression: run CollaborativeFiltering.ipynb

  - A bit more advanced logistic regression: run LogisticRegression2.ipynb

  - Simple XGBoost model: run XGBmulticlass.ipynb

  - XGBoost with feature engineering: run XGBmulticlass_withpast5.ipynb

  - Basic Neural Network: run Keras1.ipynb

  - Ensemble model: run Ensemble6_Keras1_XGB5_popular.ipynb

Special thanks to [anokas](https://www.kaggle.com/anokas) (for starter script), [BreakfastPirate](https://www.kaggle.com/breakfastpirate) (for contributing important insight to the community) and other people in the kaggle forum!