dnc1994 / kaggle-playground

Jupyter notebooks and code examples for my Kaggle Beginners' Guide.

Kaggle-Playground

After I finished the Home Depot competition, I wrote a summary as a beginners' guide to Kaggle. It got retweeted by Kaggle's official account.

In this repo you can find details about how I preprocessed the dataset, constructed features, trained different models and finally built the stacked ensemble. Main files are features.ipynb and ensemble.py.

Feel free to suggest how to improve it!

Acknowledgements

• A Wonderful Kaggle Tutorial - Andrew Conti

• Python Data Visualizations on Iris Dataset - Ben Hamner

• Beating Kaggle the Easy Way - Dong Ying

• Search Results Relevance 1st Place Winner's Interview - Chenglong Chen

• (Chinese) Solution for Prudential Life Insurance Assessment - Nutastray

Author

Linghao Zhang

License

MIT license

kaggle-playground's Issues

df_lev_dist_more_jaccard.csv问题

看到Preprocess.ipynb中有这句代码append = pd.read_csv('df_lev_dist_more_jaccard.csv', encoding='ISO-8859-1')，运行中报错，请问df_lev_dist_more_jaccard.csv这个文件从哪来的？

请问开发方式（文件结构）

我看到您的文件夹里除了ipynb 还有.py文件

请问您是如何让它们一起工作的。

在其他IDE里调试好 再复制进jupyter运行吗？

A problem in 'match_commercial'

In your features.ipynb:

In [24]:
df['match_commercial'] = (df['search_term'].str.lower().str.contains('commercial') & df['flag_commercial']).astype(np.float)
In [25]:
sum(df['match_commercial'])
Out[25]:
350.0

If df['search_term'] contains 'commercial', and df['flag_commercial'] == -1, it will return 1, is it what you expected?

Since True & (-1) will return 1.