UP my solution

sklearn_questions.py

@@ -49,9 +49,11 @@
 """
 import numpy as np
 import pandas as pd
+# import datetime as date
 
 from sklearn.base import BaseEstimator
 from sklearn.base import ClassifierMixin
+from sklearn.metrics import accuracy_score
 
 from sklearn.model_selection import BaseCrossValidator
 
@@ -82,6 +84,13 @@ def fit(self, X, y):
         self : instance of KNearestNeighbors
             The current instance of the classifier
         """
+        X, y = check_X_y(X, y)
+        check_classification_targets(y)
+        self.classes_ = np.unique(y)
+        self.X_ = X
+        self.y_ = y
+        self.n_features_in_ = X.shape[1]
+
         return self
 
     def predict(self, X):
@@ -97,7 +106,25 @@ def predict(self, X):
         y : ndarray, shape (n_test_samples,)
             Predicted class labels for each test data sample.
         """
-        y_pred = np.zeros(X.shape[0])
+        check_is_fitted(self)
+        X = check_array(X)
+
+        # calcul distances
+        distances = pairwise_distances(X, self.X_)
+
+        # select k points les plus proches
+        closest = np.argsort(distances, axis=1)[:, : self.n_neighbors]
+        closest = self.y_[closest]
+
+        # choix du y le plus frequent
+        y_pred = np.apply_along_axis(
+            lambda x: np.unique(x, return_counts=True)[0][
+                np.argmax(np.unique(x, return_counts=True)[1])
+            ],
+            axis=1,
+            arr=closest,
+        )
+
         return y_pred
 
     def score(self, X, y):
@@ -115,7 +142,9 @@ def score(self, X, y):
         score : float
             Accuracy of the model computed for the (X, y) pairs.
         """
-        return 0.
+        predictions = self.predict(X)
+
+        return accuracy_score(y, predictions)
 
 
 class MonthlySplit(BaseCrossValidator):
@@ -155,7 +184,15 @@ def get_n_splits(self, X, y=None, groups=None):
         n_splits : int
             The number of splits.
         """
-        return 0
+        X = X.reset_index()
+
+        if X[self.time_col].dtype != "datetime64[ns]":
+            raise ValueError("datetime")
+        column_date = X[self.time_col]
+        max = column_date.max()
+        min = column_date.min()
+
+        return (max.year - min.year) * 12 + max.month - min.month
 
     def split(self, X, y, groups=None):
         """Generate indices to split data into training and test set.
@@ -178,11 +215,28 @@ def split(self, X, y, groups=None):
             The testing set indices for that split.
         """
 
-        n_samples = X.shape[0]
         n_splits = self.get_n_splits(X, y, groups)
-        for i in range(n_splits):
-            idx_train = range(n_samples)
-            idx_test = range(n_samples)
-            yield (
-                idx_train, idx_test
+
+        X = X.reset_index()
+        X.index.names = ["Index_nb"]
+        X = X.reset_index()
+        X["Month"] = pd.DatetimeIndex(X[self.time_col]).month
+        X["Year"] = pd.DatetimeIndex(X[self.time_col]).year
+        X_ = X.copy()
+        X_ = X_[["Month", "Year"]].drop_duplicates().sort_values(
+            ["Year", "Month"]
             )
+        for i in range(n_splits):
+            idx_train = X.merge(
+                X_.iloc[[i]],
+                how="inner",
+                left_on=["Month", "Year"],
+                right_on=["Month", "Year"],
+            )["Index_nb"].to_numpy()
+            idx_test = X.merge(
+                X_.iloc[[i + 1]],
+                how="inner",
+                left_on=["Month", "Year"],
+                right_on=["Month", "Year"],
+            )["Index_nb"].to_numpy()
+            yield (idx_train, idx_test)