Credit Card Fraud Detection Using Deep Learning

Description

This dataset includes 2034 credit card transactions, ideal for credit card fraud detection research and development using machine learning or deep learning. It contains 31 attributes: transaction time, transaction amount, 28 variables V1-V28 which are anonymized PCA-transformed, and a binary Class label with legitimate transactions (0) and fraudulent transactions (1). The dataset can be used for anomaly detection and classification, where models for fraud detection can be tested with the help of random forest, ANN, autoencoder, and hybrid deep learning models in imbalanced data analysis.

Steps to reproduce

Obtain the original credit card transaction dataset. Remove duplicate records and check for missing values. Select the transaction features, including Time, Amount, V1–V28, and Class. Standardize the Time and Amount features where required. Apply the defined sampling/preprocessing procedure to obtain the final dataset. Save the processed records as a CSV file for machine learning and fraud detection experiments.

Institutions

• Daffodil International University

  Dhaka Division, Dhaka

Categories

Licence