Custom Mechanism Demo

This notebook demonstrates how to use a custom missing data mechanism in MissMecha.

You will learn how to:

• Define a custom masking class

• Inject it into MissMechaGenerator

• Visualize the result

# Import required modules
import numpy as np
import pandas as pd
from missmecha.generator import MissMechaGenerator
import matplotlib.pyplot as plt

np.random.seed(0)
X = pd.DataFrame({
    'feature1': np.random.randn(100),
    'feature2': np.random.rand(100) * 10,
    'target': np.random.choice([0, 1], size=100)
})
X.head()

	feature1	feature2	target
0	1.764052	4.238550	1
1	0.400157	6.063932	0
2	0.978738	0.191932	0
3	2.240893	3.015748	1
4	1.867558	6.601735	0

import numpy as np
from missmecha.generator import MissMechaGenerator

class MyCustomMasker:
    """
    A simple custom mechanism that masks the first `missing_rate` proportion of rows.
    """
    def __init__(self, missing_rate=0.1, seed=42):
        self.missing_rate = missing_rate
        self.seed = seed

    def fit(self, X, y=None):
        self.n_rows = X.shape[0]
        return self

    def transform(self, X):
        X_missing = X.astype(float).copy()
        cutoff = int(self.missing_rate * self.n_rows)
        X_missing[:cutoff, :] = np.nan
        return X_missing

# Use MissMechaGenerator with the custom mechanism
gen = MissMechaGenerator(mechanism='custom', custom_class=MyCustomMasker, missing_rate=0.2)
X_missing = gen.fit_transform(X)

compute_missing_rate(X_missing)

Overall missing rate: 20.00%
60 / 300 total values are missing.

Top variables by missing rate:

	n_missing	missing_rate (%)	n_unique	dtype	n_total
column
feature1	20	20.0	80	float64	100
feature2	20	20.0	80	float64	100
target	20	20.0	2	float64	100

{'report':           n_missing  missing_rate (%)  n_unique    dtype  n_total
 column
 feature1         20              20.0        80  float64      100
 feature2         20              20.0        80  float64      100
 target           20              20.0         2  float64      100,
 'overall_missing_rate': 20.0}