Fraud-Detection

Data obtained from https://www.kaggle.com/vbinh002/fraud-ecommerce

• Highly imbalanced data
• Time related raw data and aggregates are important/predictive of fraud

Simple RandomForest:

• accuracy_score is: : 0.9948692007515537
• roc_auc_score is: : 0.7576728137888975
• f1_score is: : 0.6712962962962962
• recall = 0.5052264808362369
• precision = 1.0

Simple RandomForest with Smote Sampling:

• accuracy_score_sm is: : 0.9948330683624801
• roc_auc_score_sm is: : 0.7641708827127756
• f1_score_sm is: : 0.6697459584295612
• recall or sens_sm = 0.5052264808362369
• precision_sm = 0.9931506849315068

Simple XGBoost with Smote Sampling:

• accuracy_score_sm is: : 0.9947969359734066
• roc_auc_score_sm is: : 0.7766622272503662
• f1_score_sm is: : 0.6682027649769584
• recall or sens_sm = 0.5052264808362369
• precision_sm = 0.9863945578231292