Fix 13 : Create SimpleImputer

diffprivlib/SimpleImputer.py

@@ -0,0 +1,56 @@
+import numpy as np
+from sklearn.base import BaseEstimator, TransformerMixin
+from diffprivlib.tools.utils import nanmean
+
+class DifferentiallyPrivateImputer(BaseEstimator, TransformerMixin):
+    def __init__(self, missing_values=np.nan, strategy='mean', fill_value=None, epsilon=1.0):
+        self.missing_values = missing_values
+        self.strategy = strategy
+        self.fill_value = fill_value
+        self.epsilon = epsilon  # Privacy budget
+
+    def fit(self, X, y=None):
+        if self.strategy == 'most_frequent':
+            self.statistics_ = [np.nan] * X.shape[1]  # Not used in most_frequent strategy
+        elif self.strategy == 'constant':
+            self.statistics_ = [self.fill_value] * X.shape[1]
+        else:
+            self.statistics_ = [nanmean(col) if np.issubdtype(col.dtype, np.number) else np.nan for col in X.T]
+        return self
+
+    def _impute_mean(self, col, missing_col):
+        non_missing_values = col[~missing_col]
+        col_mean = nanmean(non_missing_values)
+        sensitivity = np.nanmax(np.abs(non_missing_values - col_mean))
+        return col_mean, sensitivity
+
+    def _impute_median(self, col, missing_col):
+        non_missing_values = col[~missing_col]
+        col_median = np.nanmedian(non_missing_values)
+        sensitivity = np.nanmax(np.abs(non_missing_values - col_median))
+        return col_median, sensitivity
+
+    def transform(self, X):
+        noisy_X = np.copy(X)
+        missing_indices = np.isnan(X)
+
+        for col_idx in range(X.shape[1]):
+            col = X[:, col_idx]
+            missing_col = missing_indices[:, col_idx]
+
+            if np.any(missing_col):
+                if self.strategy == 'mean':
+                    col_mean, sensitivity = self._impute_mean(col, missing_col)
+                elif self.strategy == 'median':
+                    col_mean, sensitivity = self._impute_median(col, missing_col)
+                elif self.strategy == 'most_frequent':
+                    col_mean = self.statistics_[col_idx]  # Use most frequent value
+                    sensitivity = 1  # Sensitivity for most_frequent is 1
+
+                scale = sensitivity / self.epsilon
+                laplace_noise = np.random.laplace(loc=0, scale=scale, size=np.sum(missing_col))
+
+                # Impute missing values with noisy values
+                noisy_X[missing_col, col_idx] = col_mean + laplace_noise
+
+        return noisy_X