Missing Values.md

Missing Values

Several strategies for dealing with missing data:

• Eliminate data objects or attributes

• Estimate missing values

• Ignore the missing value during analysis

  Many data mining approaches can be modified to ignore missing values.

Finding missing values

df.dtypes

df.isnull().sum()

Imputation of missing values

6.4. Imputation of missing values — scikit-learn 1.0.2 documentation

One type of imputation algorithm is univariate, which imputes values in the i-th feature dimension using only non-missing values in that feature dimension (e.g. impute.SimpleImputer). By contrast, multivariate imputation algorithms use the entire set of available feature dimensions to estimate the missing values (e.g. impute.IterativeImputer).

Univariate feature imputation

sklearn.impute.SimpleImputer — scikit-learn 1.0.2 documentation

imp = SimpleImputer(missing_values=np.nan, strategy='mean')
imp.fit(
    [[1, np.nan, np.nan],
     [3, 4, np.nan]]
    )
imp.statistics_

array([ 2.,  4., nan])

可如果我只想 fit 一个列，我该怎么把列转成 fit 要求的样子？

df[[col]]

对不起，是我的想象力拖了后腿

strategy 可以为：

• mean
• median
• most_frequent
• constant (fill_value)

因为方差就是样本减去均值的平方之和，mean 显然具有最小方差。

插入的值会影响均值，进而对方差产生影响。

不依赖其它 feature 时，mean 插值具有最小的方差吗？

最优插值法_百度百科

Kriging - Wikipedia

Multivariate feature imputation

from sklearn.experimental import enable_iterative_imputer
from sklearn.impute import IterativeImputer
imp = IterativeImputer(max_iter=10, random_state=0)

请教一个深度学习的问题 - V2EX