Source code for graforvfl.shared.scaler
#!/usr/bin/env python
# Created by "Thieu" at 12:36, 17/09/2023 ----------%
# Email: nguyenthieu2102@gmail.com %
# Github: https://github.com/thieu1995 %
# --------------------------------------------------%
import numpy as np
from scipy.stats import boxcox, yeojohnson
from scipy.special import inv_boxcox
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.preprocessing import StandardScaler, MinMaxScaler, MaxAbsScaler, RobustScaler, OneHotEncoder
[docs]class LabelEncoder:
"""
Encode categorical features as integer labels.
"""
def __init__(self):
self.unique_labels = None
self.label_to_index = {}
[docs] def fit(self, y):
"""
Fit label encoder to a given set of labels.
Parameters:
-----------
y : array-like
Labels to encode.
"""
self.unique_labels = np.unique(y)
self.label_to_index = {label: i for i, label in enumerate(self.unique_labels)}
[docs] def transform(self, y):
"""
Transform labels to encoded integer labels.
Parameters:
-----------
y : array-like
Labels to encode.
Returns:
--------
encoded_labels : array-like
Encoded integer labels.
"""
if self.unique_labels is None:
raise ValueError("Label encoder has not been fit yet.")
return np.array([self.label_to_index[label] for label in y])
[docs] def fit_transform(self, y):
"""Fit label encoder and return encoded labels.
Parameters
----------
y : array-like of shape (n_samples,)
Target values.
Returns
-------
y : array-like of shape (n_samples,)
Encoded labels.
"""
self.fit(y)
return self.transform(y)
[docs] def inverse_transform(self, y):
"""
Transform integer labels to original labels.
Parameters:
-----------
y : array-like
Encoded integer labels.
Returns:
--------
original_labels : array-like
Original labels.
"""
if self.unique_labels is None:
raise ValueError("Label encoder has not been fit yet.")
return np.array([self.unique_labels[i] if i in self.label_to_index.values() else "unknown" for i in y])
[docs]class ObjectiveScaler:
"""
For label scaler in classification (binary and multiple classification)
"""
def __init__(self, obj_name="sigmoid", ohe_scaler=None):
"""
ohe_scaler: Need to be an instance of One-Hot-Encoder for softmax scaler (multiple classification problem)
"""
self.obj_name = obj_name
self.ohe_scaler = ohe_scaler
[docs] def transform(self, data):
if self.obj_name == "sigmoid" or self.obj_name == "self":
return data
elif self.obj_name == "hinge":
data = np.squeeze(np.array(data))
data[np.where(data == 0)] = -1
return data
elif self.obj_name == "softmax":
data = self.ohe_scaler.transform(np.reshape(data, (-1, 1)))
return data
[docs] def inverse_transform(self, data):
if self.obj_name == "sigmoid":
data = np.squeeze(np.array(data))
data = np.rint(data).astype(int)
elif self.obj_name == "hinge":
data = np.squeeze(np.array(data))
data = np.ceil(data).astype(int)
data[np.where(data == -1)] = 0
elif self.obj_name == "softmax":
data = np.squeeze(np.array(data))
data = np.argmax(data, axis=1)
return data
[docs]class Log1pScaler(BaseEstimator, TransformerMixin):
[docs] def fit(self, X, y=None):
# LogETransformer doesn't require fitting, so we simply return self.
return self
[docs] def transform(self, X):
# Apply the natural logarithm to each element of the input data
return np.log1p(X)
[docs] def inverse_transform(self, X):
# Apply the exponential function to reverse the logarithmic transformation
return np.expm1(X)
[docs]class LogeScaler(BaseEstimator, TransformerMixin):
[docs] def fit(self, X, y=None):
# LogETransformer doesn't require fitting, so we simply return self.
return self
[docs] def transform(self, X):
# Apply the natural logarithm (base e) to each element of the input data
return np.log(X)
[docs] def inverse_transform(self, X):
# Apply the exponential function to reverse the logarithmic transformation
return np.exp(X)
[docs]class SqrtScaler(BaseEstimator, TransformerMixin):
[docs] def fit(self, X, y=None):
# SqrtScaler doesn't require fitting, so we simply return self.
return self
[docs] def transform(self, X):
# Apply the square root transformation to each element of the input data
return np.sqrt(X)
[docs] def inverse_transform(self, X):
# Apply the square of each element to reverse the square root transformation
return X ** 2
[docs]class BoxCoxScaler(BaseEstimator, TransformerMixin):
def __init__(self, lmbda=None):
self.lmbda = lmbda
[docs] def fit(self, X, y=None):
# Estimate the lambda parameter from the data if not provided
if self.lmbda is None:
_, self.lmbda = boxcox(X.flatten())
return self
[docs] def transform(self, X):
# Apply the Box-Cox transformation to the data
X_new = boxcox(X.flatten(), lmbda=self.lmbda)
return X_new.reshape(X.shape)
[docs] def inverse_transform(self, X):
# Inverse transform using the original lambda parameter
return inv_boxcox(X, self.lmbda)
[docs]class YeoJohnsonScaler(BaseEstimator, TransformerMixin):
def __init__(self, lmbda=None):
self.lmbda = lmbda
[docs] def fit(self, X, y=None):
# Estimate the lambda parameter from the data if not provided
if self.lmbda is None:
_, self.lmbda = yeojohnson(X.flatten())
return self
[docs] def transform(self, X):
# Apply the Yeo-Johnson transformation to the data
X_new = boxcox(X.flatten(), lmbda=self.lmbda)
return X_new.reshape(X.shape)
[docs] def inverse_transform(self, X):
# Inverse transform using the original lambda parameter
return inv_boxcox(X, self.lmbda)
[docs]class SinhArcSinhScaler(BaseEstimator, TransformerMixin):
# https://stats.stackexchange.com/questions/43482/transformation-to-increase-kurtosis-and-skewness-of-normal-r-v
def __init__(self, epsilon=0.1, delta=1.0):
self.epsilon = epsilon
self.delta = delta