使用 Python 创建数据集的可编辑副本，同时保持原始数据不变

Question

I have recently started studying machine learning and am focusing on the pre-processing stage. I'm creating a Jupyter Notebook which lays out the stages of pre-processing step by step. The original dataset has some missing values in it which I will replace with the mean value. As it is a step by step notebook I am creating, I would like to be able to keep the original dataset in tact while having a copy of the dataset which will be updated at different steps of the process ie missing cells will be replaced with the mean values. See the code below for what I have done so far. It does what I want it to do so far, it's just missing the copied dataset part.

Any tips or links to tutorials would be appreciated. Thanks.

#libraries
import matplotlib.pyplot as plt
import numpy as np 
import pandas as pd 

from sklearn.impute import SimpleImputer

#Importing dataset
dataset = pd.read_csv('example.csv')

# Splitting the attributes into independent and dependent attributes
X = dataset.iloc[:, :-1].values # attributes to determine dependent variable
Y = dataset.iloc[:, 4].values # dependent variable / Class variable, final column


#printing and displaying dataset
print(dataset)
display(dataset.describe())

#check how many null values in dataset and output value
print('Number of null/NaN values in dataset: ',dataset.isnull().values.sum())

#show how many null values per column
print(dataset.isnull().sum())

#DEALING WITH MISSING VALUES USING MEAN
from sklearn.impute import SimpleImputer

#creating SimpleImputer object, specifying to change missing values to mean
imputer = SimpleImputer(missing_values=np.nan, strategy='mean')

#MEAN
imputer = imputer.fit(X[:, 3:4])
X[:,3:4] = imputer.transform(X[:,3:4])

print(X)

Answer 1

I would rather prefer to use copy class in python by calling deepcopy

import copy
df_edit = copy.deepcopy(df_original)

Now you can play with df_edit and make your changes without disturbing df_original

or

you can directly use pandas copy module like df_edit = df_original.copy(deep = True)

Answer 2

So you just want to create a copy of dataset to revert to later?

You could create a chunk in your notebook to back it up eg

dataset_backup = dataset

then another to overwrite dataset with dataset backup

dataset = dataset_backup

Then run each as and when you want to backup the dataset or revert to the backup

Alternatively, if you want to have a record of dataset at each step in the process just create a new variable for it each time eg dataset_means this is usually a good idea for debugging purposes.

Is this what you were asking?