Python-sklearn-Normalization

• 六种常见归一化方法的适用场景和映射公式

"""
@Title: Normalization
@Time: 2024/2/21
@Author: Michael Jie
"""

import numpy as np
from sklearn.preprocessing import MinMaxScaler, StandardScaler

"""
线性归一化、均值归一化：
1、最大最小值明确不变
2、数据相对稳定
3、对处理后的数据范围有严格要求

零-均值归一化：
1、在数据存在异常值、最大最小值不固定

非线性归一化（对数归一化、反正切函数归一化、小数定标标准化）：
1、数据分化程度较大，需要数学函数对原始值进行映射
"""


# 线性归一化，将数据值映射到[0, 1]之间，适用于数据较集中的情况
def max_min_normalization(x):
    x = np.array(x)
    x = (x - np.min(x)) / (np.max(x) - np.min(x))
    return list(x)


# 均值归一化
def mean_normalization(x):
    x = np.array(x)
    x = (x - np.mean(x)) / (np.max(x) - np.min(x))
    return list(x)


# 零-均值归一化（标准化）
def z_score_normalization(x):
    x = np.array(x)
    x = (x - np.mean(x)) / np.std(x)
    return list(x)


# 对数归一化
def log_normalization(x):
    x = np.array(x)
    x = np.log(x) / np.log(np.max(x))
    return list(x)


# 反正切函数归一化
def arctan_normalization(x):
    x = np.array(x)
    x = np.arctan(x) * (2 / np.pi)
    return list(x)


# 小数定标标准化
def decimal_normalization(x):
    x = np.array(x)
    k = 0
    x = x / (10 ** k)
    return list(x)


if __name__ == '__main__':
    datas = [1, 0.7, 2.4, 2.1, 1.5, 2.7, 0.9, 1.6, 0.5, 10]
    # 数据不稳定，使用线性归一化，数据映射趋近于0
    print("线性归一化：{x}".format(x=max_min_normalization(datas)))
    print("均值归一化：{x}".format(x=mean_normalization(datas)))
    print("零-均值归一化（标准化）：{x}".format(x=z_score_normalization(datas)))

    # 线性归一化
    min_max_scaler = MinMaxScaler()  # 实例化
    result = min_max_scaler.fit_transform(datas)  # 归一化
    min_max_scaler.inverse_transform(result)  # 归一化逆操作

    # 零-均值归一化（标准化）
    standard_scaler = StandardScaler()
    result = standard_scaler.fit_transform(datas)
    standard_scaler.inverse_transform(result)

    """
    线性归一化：[0.05263157894736842, 0.021052631578947364, 0.19999999999999998, 0.16842105263157894, 0.10526315789473684,
    0.23157894736842108, 0.042105263157894736, 0.11578947368421054, 0.0, 1.0]
    均值归一化：[-0.1410526315784735, -0.172631578947684, 0.006315789736842165, -0.02526315789736817, -0.0884210263157893,
    0.037894736842105294, -0.15157894736842104, -0.07789473684210524, -0.19368421052631576, 0.8063157894736842]
    零-均值归一化（标准化）：[-0.5062410229091, -0.6195785499784, 0.02266750858865, -0.0907003336553, -0.313451184177938, 
    0.1360050507504832, -0.5440202030019322, -0.2795659376537707, -0.6951369260580246, 2.893885246524168]
    """