The goal of this project is to detect anomalies presented in Online payment transactions and classifies transactions as fraudulent and not.
- Statistically and Analytically analyzed the data and using BenfordsLaw detected the presence of anomalies in the data.
- Compared different anomaly detection models based on the anomaly detection performance and separated anomalies presented using iForest Algorithm.
- Utilizing Classification algorithms, classified fraudulent transactions from non-fraudulent transactions for future detection of frauds. Acquired Classification Accuracy of 99% with LighGBM Algorithm.
- Created Preprocessing including balancing data using SMOTE and modeling Pipeline using Pycaret Opensource Library.
- Transferred models and pipeline to Tableau for detecting anomalies in the future data and to Build a dashboard using Tableau.
- Python Packages: Pandas, Seaborn, Matplotlib, Pycaret
- Anomaly Detection Model : iForest,knn,SVM,Cluster
- Classification Model : LightGBM
- Tableau