arjunssat / credit_card_fraud_detection

Detection of Credit_card flaws and analysis through various models

Credit Card Fraud Detection

What is Credit Card Fraud Detection?



Credit card fraud is a term that has been coined for unauthorized access of payment cards like credit cards or debit cards to pay for using services or goods Hacker's or fraudster's may obtain the confidential details of the card from unsecured websites. When a fraudster compromises an individual's credit/debit card, everyone involved in the process suffers, right from the individual whose confidential data has been leaked to the businesses (generally banks) who issue the credit card and the merchant who is finalizing the transaction with purchase.

This makes it extremely essential to identify the fraudulent transactions at the onset. Financial institutions and businesses like e-commerce are taking firm steps to flag the fraudsters entering the system. Various advanced machine learning technologies are at play, assessing every transaction and stemming the fraud users in its nip using behavioral data and transaction pattern. The process of automatically differentiating between fraudulent and genuine users is known as “credit card fraud detection”

Credit Card Fraud Detection's using Machine Learning can be done using:


• Unsupervised Learning
-
* Machine Learning Algorithms such as Isolation Forest, One-class SVM, LOF, etc., do not require labeled data for training the model. They identify patterns in the data and try to group the data points based on observed similarities in patterns.


• Supervised Learning
-
* Machine Learning and Deep Learning Algorithms such as Ensemble Models (RandomForest, XGBoost, LightGBM, etc.), KNN, Neural Networks, Autoencoders, etc. These algorithms are trained on labeled data, and the model learns to predict the labels for the unseen data. Labeled data can be expensive to gather.




Challenges in Credit Card Fraud Detection's

The challenges involved in credit card fraud detection project is primarily the data itself. The data is heavily imbalanced, i.e., the count of data labeled as fraudulent is way less than the data labeled as non-fraudulent data. This makes it extremely tricky to train the model as it tends to overfit for the majority class and underfit for the minority class. Techniques like oversampling, undersampling, cost-sensitive learning, etc. can be used to deal with this. The metrics used for the final model are different from standard evaluation metrics of accuracy, AUC-ROC, etc.

Another prevalent faced challenge is the quality and quantity of the data. The startups in the early stage do not have much user history data to train extensive models, which makes it difficult to train a robust fraud detection model. A temporary solution to this problem can be sourcing data from an external third party, like scores from credit bureaus.