Credit Card Fraud Detection

Built an anomaly detection pipeline to identify fraudulent credit card transactions on a highly imbalanced dataset. Compared Local Outlier Factor (LOF) and Isolation Forest to flag transactions that deviate from normal spending patterns.

Tools: Python, scikit-learn, LOF, Isolation Forest
Repo: GitHub

---

Key Visualizations

Transaction class distribution

The dataset is heavily skewed toward legitimate transactions, making class imbalance a core challenge.

Exploratory data analysis

Visual exploration of feature distributions to understand separation between normal and fraudulent transactions.

Correlation matrix

Feature correlation analysis to identify redundant variables and inform model selection.

---

Approach

1. Explored class imbalance and transaction feature distributions
2. Applied LOF to score anomaly likelihood per transaction
3. Used Isolation Forest as a complementary unsupervised detector
4. Evaluated detection performance on the imbalanced dataset

---

Links

• GitHub repository