Credit Card Fraud Detection using Machine Learning methods

Introduction

In this project, the aim is to import and use several ML models on Credit Card Fraud Detection dataset (on Kaggle) in order to classify fraudulent and non-fraudulent transactions and compare their performances based on how well they can distinguish fraudulent transactions, rather than merely calculating accuracy-scores.

Since only 0.17% of the transaction records in this dataset are fraudulent, i.e. the dataset is highly skewed, it is not plausible to adopt accuracy-score as the evaluation metric. Although, this metric is calculated, other metrics, such as precision and f1-score, are also used. In order to address the skewness of the dataset, undersampling technique is adopted.

In this project, the following ML models are used: Decision Tree, Random Forest, Logistic Regression, Gaussian Naive Bayes, KNN, XGBoost, AdaBoost, and a model ensemble.

The ensemble adopts Majority Voting technique. When creating the ensemble, in 30 iterations, a weighted combination of the mentioned models is created and then, their majority vote is taken. The combination that is created in each iteration consists of 2 to 5 randomly chosen models from a list, sorted by f1-scores.

The evaluation metrics corresponding to each model, or combination of models, is printed.

Finally, the Precision-Recall and ROC Curves of all models are plotted for more convinience.

Installation

First, you need to clone this repository to your local machine via the following command:

$ git clone https://github.com/aylinghsr/Fraud_Detection.git

In case you don't have git installed on your computer, you can download the zip file of this repository and then, extract it.

Requirements

This project is written in Python3 and requires Scikit-learn, Pandas, and Numpy libraries.

All the required libraries can be installed by running the following command:

$ pip install -r requirements.txt

If the command above results in an error, you can also try:

$ python -m pip install -r requirements.txt

Also, the dataset (.csv file) should be downloaded on your computer.

Dataset: Credit Card Fraud Detection

Usage

Run:

$ cd Fraud_Detection
$ python Fraud_Detection.py