Python：基于类对numy array类型数据预处理的实现（标准化、归一化、均值化、min-max等）

        将自己写过的一些代码和功能整理一下，数据科学这块首先想做的就是关于数据的预处理，通过类的方式可以比较轻松地整合诸多数据预处理的方法，以下为源码。

import numpy as np
import pandas as pd

"""This file is used to do data pre-processing.
The class DataPreProcessingForNumpyArray contains functions ad shown below:
standarlize:标准化(Z-score)
noemalization:0-1化
equalization:均值化
min_max_negation:min-max负值化"""

class DataProcessingForNumpyArray:
    def __init__(self,data):
        """To store data for being processed"""
        self.data = data
        (self.r, self.c) = self.data.shape

    """Standardlize the data"""
    def standarlization(self, r_or_c = 0, selected = None):
        """Arguments:
        r_or_c: int determines whether the function processes data by rows or columns,
            0 refers to rows and 1 refers to columns, 0 is the default value
        selected: list determines which rows or columns to be processed
            if not defiend, the function will process all data
        """
        def Z_Score(data):
            mean = np.mean(data)
            std_dev = np.std(data)
            norm_data = (data - mean) / std_dev
            return norm_data
        if selected == None:
            selected = [i for i in range(0,self.data.shape[r_or_c])]
        if r_or_c == 0:
            for i in selected:
                data = self.data[i,:]
                self.data[i,:] = Z_Score(data)
        elif r_or_c == 1:
            for j in selected:
                data = self.data[:,j]
                self.data[:,j] = Z_Score(data)

    """Normalize the data"""
    def normalization(self, r_or_c = 0, selected = None):
        """Arguments:
        r_or_c: int determines whether the function processes data by rows or columns,
            0 refers to rows and 1 refers to columns, 0 is the default value
        selected: list determines which rows or columns to be processed
            if not defiend, the function will process all data
        """
        def unil(data):
            data = (data - np.min(data)) / (np.max(data) - np.min(data))
            return data
        if selected == None:
            selected = [i for i in range(0,self.data.shape[r_or_c])]
        if r_or_c == 0:
            for i in selected:
                data = self.data[i,:]
                self.data[i,:] = unil(data)
        elif r_or_c == 1:
            for j in selected:
                data = self.data[:,j]
                self.data[:,j] = unil(data)

    """Equalize the data"""
    def equalization(self, r_or_c = 0, selected = None):
        """Arguments:
        r_or_c: int determines whether the function processes data by rows or columns,
            0 refers to rows and 1 refers to columns, 0 is the default value
        selected: list determines which rows or columns to be processed
            if not defiend, the function will process all data
        """
        if selected == None:
            selected = [i for i in range(0,self.data.shape[r_or_c])]
        if r_or_c == 0:
            for i in selected:
                data = self.data[i,:]
                self.data[i,:] = self.data[i,:]/np.mean(data)
        elif r_or_c == 1:
            for j in selected:
                data = self.data[:,j]
                self.data[:,j] = self.data[:,j]/np.mean(data)

    """Negatize """
    def min_max_negation(self,r_or_c =0,selected = None):
        """Arguments:
        r_or_c: int determines whether the function processes data by rows or columns,
            0 refers to rows and 1 refers to columns, 0 is the default value
        selected: list determines which rows or columns to be processed
            if not defiend, the function will process all data
        """
        if selected == None:
            selected = [i for i in range(0, self.data.shape[r_or_c])]
        if r_or_c == 0:
            for i in selected:
                data = self.data[i,:]
                ma = np.max(data)
                mi = np.min(data)
                self.data[i,:] = ma - data + mi
        elif r_or_c == 1:
            for j in selected:
                data = self.data[:,j]
                self.data[:,j] = self.data[:,j]/np.mean(data)

    """Save data as an excel file"""
    def save_data(self,filename,savepath = None):
        """filename: a string end with ".xlsx"
        savepath: an availible system path"""
        df = pd.DataFrame(self.data)
        if not savepath:
            df.to_excel(filename)
        else:
            df.to_excel(savepath + filename)

        该类一共有五个方法——四个是数据预处理的方法，一个是用于保存数据，四种预处理方法的具体功能根据其名字就能看出，具体使用方法是先实例化该类，然后调用方法，如果需要的话可以保存。

第一步，实例化该类需要传入type为numpy array且元素类型均为int或float的对象。

第二步，调用该实例下的四种预处理方法，它们有一致的参数：r_or_c为0或1的整型数据，指代是以行或列处理数据，0代表将一行视为一个指标进行处理，1代表将一列视为一个指标进行处理；selected应当是list，是选填选项，指代需要处理的行或列，若不指定则默认处理所有行或列。

第三步，保存数据必填参数为以".xlsx"为结尾的文件名，选填保存路径，若不填保存路径则默认保存在程序所在的文件夹下。

使用例，其中，”data.xlsx“为如下图所示的数据表格：

save_path = 'C:\\Users\\Technicolor\\Desktop\\pythonProject\\test\\datas\\'
Data = pd.read_excel("data.xlsx")
data = Data.values[:,1:]
dt = DataProcessingForNumpyArray(data)
dt.equalization(r_or_c=1)
dt.save_data(filename="test.xlsx",savepath=save_path)