chugh007/CreditCard_FraudDetection

Credit Card Fraud Detection using Anomaly Detection

Jupyter Notebook

CreditCard_FraudDetection

General Information

Dataset : https://www.kaggle.com/mlg-ulb/creditcardfraud
Run in Google Collab : Credit Card Fraud
Metric Used : Area Under the ROC curve
Technique : Anomaly Detection

This is a classic example to practice anomaly detection . I have followed the steps from Andrew NG's machine learning tutorial for anomaly detection (https://youtu.be/086OcT-5DYI)

Implementation Notes

As mentioned in the tutorial , the features used follows gaussian distribution.
To estimate the probability density at a point ,GaussianMixture is used .
Training set has all non-anomolous samples
The anomolous samples are split equally among test and validation sets.
The optimal threshold value for highest roc_auc_score is chosen using validation set.

Future Works

AIC and BIC score is used to estimate the number of clusters . We can also use bayesian mixture model which can give a good estimate of the number of clusters, but it is very slow to train.