The project consists of a highly imbalanced target class. Our goal is to recognize fraudulent credit card transactions so that customers are not charged for purchases they did not make. We deploy ensemble methods and boosting algorithms like Random Forest, Xgboost, and LightGBM. We further try to improve our prediction using a Voting classifier and stacking techniques to get a better AUC score.
The data consist of 30 feature variables and 1 target variable with 284,806 instances, provided by Machine Learning Group on Kaggle
Our analysis include following topics (for detailed analysis click the link)
| page_link | notebook | |
|---|---|---|
| Data Processing | main | notebook |
| Exploratory Data Analysis | main | notebook |
| Balanced Data | main | notebook |
| Random Forest Classifier | main | notebook |
| Xgboost | main | notebook |
| LightGBM | main | notebook |
| Stacking_Voting | main | notebook |
