Fraud Credit Card Detection - Group 17

Overview

This project is focused on the detection of fraudulent activities in credit card transactions. We utilize various machine learning models to classify transactions as either fraudulent or legitimate based on transactional data.

The primary objective of this project is to enhance the security of credit card transactions. By accurately identifying fraudulent transactions, we can prevent unauthorized activities and safeguard the financial assets of cardholders.

Models Used

In this project, we tried to implement and evaluate the performance of the following machine learning models:

• Logistic Regression
• Decision Trees
• Random Forest
• Support Vector Machine
• K-nearest neighbors
• Multi-layer Perception
• Deep Neural Networks

Results

Here is th summary of the results we obtained from the models:

Methods	Accuracy	Precision	F1-Score	Recall
LR	0.96	0.93	0.96	0.98
SVM	0.95	0.95	0.94	0.95
DT	1.0	1.0	1.0	1.0
RF	1.0	1.0	1.0	1.0
KNN	0.95	0.98	0.91	0.84
MLP	0.99	0.64	0.69	0.75
DNN	0.89	0.90	0.82	0.75

Run the project

1. Clone the Repository

   Clone this repository to your local machine using the following command:

   git clone https://github.com/chutrunganh/Fraud-Credit-Card-Detection-Group-17.git

2. Install Dependencies

   Check the requirement.txt file. It lists all Python libraries required for running the code. Ensure that you have Python 3.6 or higher installed.

3. Dataset

   All the Datasets we used located in the Dataset directory. Due to different paths depending on the environment, you may need to change the path to the dataset in the source code.

4. Run the Code

• For Dataset Characteristics Analysis:

  • For Dataset1 and Dataset2, refer to Data Characteristic.ipynb.
  • For Dataset3's characteristics and preprocessing steps, see DataPreprocessingForDataset3.ipynb

• For Model Evaluation: Consult RunModels.ipynb to review the outcomes from all models. Specifically, we have implemented our own version of the KNN model in Java, available in the KNN\src directory.

• If you want to run this on Google Colab, access the source code via this link: Google Drive.

Some Notes

Ensure that the version of the scikit-learn == 1.2.2, while running in google colab, it might return no errors, but there might be error on local machine due to incompatible version with SMOTE library we use, detail here: https://stackoverflow.com/questions/76593906/how-to-resolve-cannot-import-name-missingvalues-from-sklearn-utils-param-v

Also, for testing purposes, we strongly recommend choosing Undersampling technique module in the source code as it significantly reduces the runtime (With some models like MLP, DNN, SVM can take more than 30 min for each hyperparameter tuning part with SMOTE technique). However, please note that Undersampling may not provide reliable evaluation results.

Documentation

Detailed about this project can be found in our Report in the Docs directory.

Contributors

This project was developed by Group 17 for the Machine Learning course IT3910E, semester 20232 at Ha Noi University of Science and Technology.

Name	ID
Vũ Hoàng Nhật Anh (Leader)	20225471
Chu Trung Anh	20225564
Trần Nhật Minh	20225511
Trần Nam Tuấn Vượng	20225540