Скачать или смотреть #22: Scikit-learn 19: Preprocessing 19: Compare imputation techniques

The video discusses the code and results from different imputation techniques in Scikit-learn in Python.

Timeline
(Python 3.8)

00:00 - Outline of video
00:39 - Open Jupyter notebook
01:22 - Data
03:34 - Mean squared error (MSE)
05:58 - Initialize empty arrays for MSE's and Standard Deviation's
06:30 - Full score: Regressor: RandomForestRegressor()
09:20 - Zero impute: SimpleImputer()
10:36 - -- CORRECTION -- meant to say: "three of these extra columns attached..." and not rows attached
14:51 - KNN impute: KNNImputer()
16:51 - Mean impute: SimpleImputer(strategy='mean')
17:39 - IterativeImputer()
19:22 - Plot MSE's
21:00 - Imputed values: Zero impute
22:30 - Imputed values: Mean impute
23:16 - Imputed values: KNN impute
24:05 - Imputed values: IterativeImputer
25:35 - Ending notes


#########
Data
#########

x = np.array([
[-1, 12, 1, 1, 1, 0],
[-2, 13, 1, 2, 2, 1],
[-3, 14, 1, 3, 4, 1],
[-4, 1, 1, 4, 8, 2],
[-5, 2, 2, 5, 16, 4],
[-6, 3, 2, 6, 32, 8],
[-7, 12, 2, 7, 64, 16],
[-8, 13, 2, 8, 128, 32]
])
y = np.sum(x, axis=1)

x_missing = np.array([
[-1, 12, 1, 1, 1, 0],
[np.nan, 13, 1, 2, 2, 1],
[-3, np.nan, 1, 3, 4, 1],
[-4, 1, np.nan, 4, 8, 2],
[-5, 2, 2, np.nan, np.nan, 4],
[-6, 3, 2, 6, 32, 8],
[-7, 12, 2, 7, 64, 16],
[-8, 13, 2, 8, 128, np.nan]
])

y_missing = y.copy()

#########