Impute Demo

This notebook demonstrates how to use the SimpleSmartImputer module from MissMecha for basic missing value imputation.

We will:

• Create a small dataset with both numerical and categorical features

• Apply automatic imputation strategies

• Inspect the imputed results

# --- Import Required Libraries ---
import pandas as pd
import numpy as np

# --- Set random seed for reproducibility ---
np.random.seed(42)

# --- Create a Synthetic Dataset ---

data = {
    'age': [25, np.nan, 30, 22, np.nan, 28, 35, 40, np.nan, 32],
    'income': [50000, 60000, np.nan, 52000, 58000, np.nan, 61000, 59000, 57000, np.nan],
    'gender': ['M', 'F', 'M', np.nan, 'F', 'F', 'M', 'F', np.nan, 'F']
}

df = pd.DataFrame(data)

print("Original dataset with missing values:")
df.head()

Original dataset with missing values:

	age	income	gender
0	25.0	50000.0	M
1	NaN	60000.0	F
2	30.0	NaN	M
3	22.0	52000.0	NaN
4	NaN	58000.0	F

# --- Apply SimpleSmartImputer ---

from missmecha.impute import SimpleSmartImputer

# Instantiate the imputer
imp = SimpleSmartImputer(cat_cols=['gender'])

# Fit and transform the data
df_imputed = imp.fit_transform(df)

print("Imputed dataset:")
df_imputed.head()

[SimpleSmartImputer] Column 'age' treated as numerical. Fill value = 30.285714285714285
[SimpleSmartImputer] Column 'income' treated as numerical. Fill value = 56714.28571428572
[SimpleSmartImputer] Column 'gender' treated as categorical. Fill value = F
Imputed dataset:

	age	income	gender
0	25.000000	50000.000000	M
1	30.285714	60000.000000	F
2	30.000000	56714.285714	M
3	22.000000	52000.000000	F
4	30.285714	58000.000000	F

Note

• Numerical columns (e.g., age, income) are imputed using the mean.

• Categorical columns (e.g., gender) are imputed using the mode.

• No manual column type specification is required; MissMecha detects types automatically.

For more advanced control, users can manually specify column types or customize imputation behavior.

Key Takeaways

• Automatically imputes numerical columns with mean and categorical columns with mode.

• Supports scikit-learn style API (fit, transform, fit_transform).

• Allows manual specification of categorical columns if needed.

• Provides a quick and lightweight baseline for missing data imputation.