sklearn の OrdinalEncoder の挙動

OrdinalEncoder の挙動についてのメモ

sklearn.preprocessing.OrdinalEncoder - scikit-learn 0.24.2 documentation

The input to this transformer should be an array-like of integers or strings, denoting the values taken on by categorical (discrete) features. The features are converted to ordinal integers. This results in a single column of integers (0 to n_categories - 1) per feature.

scikit-learn.org

カテゴリカル feature を int にします

>>> df = pd.DataFrame([1,2,3,2,1])
>>> oe = OrdinalEncoder(categories='auto', dtype=np.int64)
>>> oe.fit(df)
>>> oe.transform(df)
array([[0],
       [1],
       [2],
       [1],
       [0]])

>>> oe.categories_
[array([1, 2, 3])]

カテゴリーの型は全て同じである必要がある

必要があれば value を str に変換するなど必要

df = pd.DataFrame([1,2,3,2,1]).astype('str') # astype(str) すれば問題ない
categories = [['1', '2', '3', 'na']]
oe = OrdinalEncoder(categories=categories, dtype=np.int64)
oe.fit(df)
oe.transform(df)

>>> df = pd.DataFrame([1,2,3,2,1]) >>> oe = OrdinalEncoder(categories='auto', dtype=np.int64) >>> oe.fit(df) >>> oe.transform(df) array([[0], [1], [2], [1], [0]]) >>> oe.categories_ [array([1, 2, 3])]

>>> df = pd.DataFrame([1,2,3,2,1]) >>> categories = [[1,2,3, 'na']] >>> oe = OrdinalEncoder(categories=categories, dtype=np.int64) >>> oe.fit(df) --------------------------------------------------------------------------- ValueError Traceback (most recent call last) <ipython-input-92-fbe59670fc3d> in <module>() 2 categories = [[1,2,3, 'na']] 3 oe = OrdinalEncoder(categories=categories, dtype=np.int64) ----> 4 oe.fit(df) 5 oe.transform(df) 1 frames /usr/local/lib/python3.7/dist-packages/sklearn/preprocessing/_encoders.py in _fit(self, X, handle_unknown) 86 cats = _encode(Xi) 87 else: ---> 88 cats = np.array(self.categories[i], dtype=Xi.dtype) 89 if Xi.dtype != object: 90 if not np.all(np.sort(cats) == cats): ValueError: invalid literal for int() with base 10: 'na'

df = pd.DataFrame([1,2,3,2,1]).astype('str') # astype(str) すれば問題ない categories = [['1', '2', '3', 'na']] oe = OrdinalEncoder(categories=categories, dtype=np.int64) oe.fit(df) oe.transform(df)