snaplib 0.4.70

Data preprocessing lib

snaplib

A simple data preprocessing tools.

---

user guide

Kaggle Notebook

Classification

https://www.kaggle.com/code/artyomkolas/titanic-snaplib-classification/notebook

Regression

https://www.kaggle.com/code/artyomkolas/housing-prices-with-snaplib/notebook

---

PyPi

!pip install snaplib
from snaplib.snaplib import Snaplib
sl = Snaplib()

1. sl.nan_info
2. sl.nan_plot
3. sl.cleane
4. sl.recover_data - NaN imputing with ML
5. sl.train_test_split_balanced
6. sl.encode_dataframe
7. sl.decode_dataframe
8. sl.k_folds_split

For one and list of algorithms with bagging

9. sl.cross_val
10. sl.features_selection_regr
11. sl.features_selection_clsf
12. sl.fit_stacked
13. sl.save_stack
14. sl.load_stack
15. sl.predict_stacked

doc

print(sl.recover_data.__doc__)

Imputing of missing values (np.nan) in tabular data, not TimeSeries.

Use case: df = Snaplib().recover_data(df, device="cpu", verbose=True)
device must be "cpu" or "gpu". Sometime small datasets work faster with cpu.
verbose = True algorithm runs cross validation tests and prints results of tests for decision making.
discrete_columns = ['col_name_1', 'col_name_2', 'col_name_3', 'etc']

TESTS on https://www.kaggle.com/code/artyomkolas/nan-prediction-in-progress/notebook