/Fraud-Detection-using-Neural-Network

Primary LanguageJupyter Notebook

Fraud Detection Using Deep Learning

Predicting fraud is a slightly different case, when designing a predictive algorithm. 4 algorithms are implemented here using under sampling technique while using k-fold cross validation.

Accuracy

Logistic regression - 65.99%

Neural network - 68.75 %

Xgboost - 75.66 %

Xgboost with unstratified splitting - 85.84 %

The script uses: Python version=3.5.5, Keras=2.2.0, TensorFlow=1.8.0.