Python Implementation of Resampling Techniques
We will apply both undersampling and oversampling our dataset for balancing our target variable
Step 1: Import Libraries
Python3
import pandas as pd import numpy as np import seaborn as sns from sklearn.preprocessing import StandardScaler from imblearn.under_sampling import RandomUnderSampler, TomekLinks from imblearn.over_sampling import RandomOverSampler, SMOTE |
Step 2: Reading the Dataset
We will read the dataset using the pandas read_csv function. Also, we will see the percentage of each class in the target variable of the dataset.
Python3
dataset = pd.read_csv('creditcard.csv') print("The Number of Samples in the dataset: ", len(dataset)) print('Class 0 :', round(dataset['Class'].value_counts()[0] / len(dataset) * 100, 2), '% of the dataset') print('Class 1(Fraud) :', round(dataset['Class'].value_counts()[1] / len(dataset) * 100, 2), '% of the dataset') |
Output:
The Number of Samples in the dataset: 284807 Class 0 : 99.83 % of the dataset Class 1(Fraud) : 0.17 % of the dataset
Step3: Undersampling of Major class
We will do an undersampling of the major class where the cardholder is not a fraud through this technique we will reduce the rows which come under the major class.
Python3
X_data = dataset.iloc[:, :-1] Y_data = dataset.iloc[:, -1:] rus = RandomUnderSampler(random_state=42) X_res, y_res = rus.fit_resample(X_data, Y_data) X_res = pd.DataFrame(X_res) Y_res = pd.DataFrame(y_res) print("After Under Sampling Of Major Class Total Samples are :", len(Y_res)) print('Class 0 :', round(Y_res.value_counts()\ [0] / len(Y_res) * 100, 2), '% of the dataset') print('Class 1(Fraud) :', round(Y_res.value_counts()\ [1] / len(Y_res) * 100, 2), '% of the dataset') |
Output:
After Under Sampling Of Major Class Total Samples are : 984 Class 0 : 50.0 % of the dataset Class 1(Fraud) : 50.0 % of the dataset
We can see after doing undersampling the total data in the major class has reduced to 984.
Step4: Undersampling Using Tomelinks
We can do undersampling using the Tomelinks library.
Python3
tl = TomekLinks() X_res, y_res = tl.fit_resample(X_data, Y_data) X_res = pd.DataFrame(X_res) Y_res = pd.DataFrame(y_res) print("After TomekLinks Under Sampling Of Major\ Class Total Samples are :", len(Y_res)) print('Class 0 :', round(Y_res.value_counts()\ [0] / len(Y_res) * 100, 2), '% of the dataset') print('Class 1(Fraud) :', round(Y_res.value_counts()\ [1] / len(Y_res) * 100, 2), '% of the dataset') |
Output:
After TomekLinks Under Sampling Of Major Class Total Samples are : 284736 Class 0 : 99.83 % of the dataset Class 1(Fraud) : 0.17 % of the dataset
Step5: Oversampling Using RandomOversampler
We can use RandomOversampler to oversample the minority class data. Using Random Oversample the model picks randomly data points from the existing datasets.
Python3
ros = RandomOverSampler(random_state=42) X_res, y_res = ros.fit_resample(X_data, Y_data) X_res = pd.DataFrame(X_res) Y_res = pd.DataFrame(y_res) print("After Over Sampling Of Minor Class\ Total Samples are :", len(Y_res)) print('Class 0 :', round(Y_res.value_counts()\ [0] / len(Y_res) * 100, 2), '% of the dataset') print('Class 1(Fraud) :', round(Y_res.value_counts()\ [1] / len(Y_res) * 100, 2), '% of the dataset') |
Output :
After Over Sampling Of Minor Class Total Samples are : 568630 Class 0 : 50.0 % of the dataset Class 1(Fraud) : 50.0 % of the dataset
Step6: Oversampling Using SMOTE
We can use Smote to generate random sample data for the minority class. One useful thing about using SMOTE is that it creates new synthetic data points for the minority class.
Python3
sm = SMOTE(random_state=42) X_res, y_res = sm.fit_resample(X_data, Y_data) X_res = pd.DataFrame(X_res) Y_res = pd.DataFrame(y_res) print("After SMOTE Over Sampling Of Minor\ Class Total Samples are :", len(Y_res)) print('Class 0 :', round(Y_res.value_counts()\ [0] / len(Y_res) * 100, 2), '% of the dataset') print('Class 1(Fraud) :', round(Y_res.value_counts()\ [1] / len(Y_res) * 100, 2), '% of the dataset') |
Output:
After SMOTE Over Sampling Of Minor Class Total Samples are : 568630 Class 0 : 50.0 % of the dataset Class 1(Fraud) : 50.0 % of the dataset
Introduction to Resampling methods
While reading about Machine Learning and Data Science we often come across a term called Imbalanced Class Distribution, which generally happens when observations in one of the classes are much higher or lower than in other classes.
As Machine Learning algorithms tend to increase accuracy by reducing the error, they do not consider the class distribution. This problem is prevalent in examples such as Fraud Detection, Anomaly Detection, Facial recognition, etc.