Anomaly-Detection-in-Financial-Transactions

Goal: To Identify Anomalous(Fraud) credit-card transactions with lowest possible False Positive Rate

Worked on a highly imbalanced dataset with 30 PCA-transformed features and a class ratio of 1000:17.
Applied Supervised Classification Algorithms on the dataset split using stratified random sampling, among which the Random Forest Classifier gave a higher Recall score of 0.78.
Improved the Recall score to 0.82 with the Isolation Forest model trained in an unsupervised setting.
Designed a Deep Auto-Encoder Neural Network model which achieved the best recall score of 0.85 with only a 6.5% False Positive Rate on the fraudulent class.

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
01-EDA-checkpoint.ipynb		01-EDA-checkpoint.ipynb
02-Preprocessing-checkpoint.ipynb		02-Preprocessing-checkpoint.ipynb
03-Modeling_ML-checkpoint.ipynb		03-Modeling_ML-checkpoint.ipynb
04-Modeling_ML-checkpoint.ipynb		04-Modeling_ML-checkpoint.ipynb
05-Auto-Encoder-checkpoint.ipynb		05-Auto-Encoder-checkpoint.ipynb
README.md		README.md

Provide feedback