Anomaly Detection performed on Credit Card Fraud Detection dataset by Kaggle (https://www.kaggle.com/mlg-ulb/creditcardfraud).
Nowadays, credit cards play a determinant role in many people's life. According to statistics, there are 2.8 Billion credit cards in use worldwide of which 1.06
Billions in the United States. Moreover, as reported by Experian State of Credit Report and the Fifth report on credit fraud published by the European Central Bank, 67% of the American people own at least one credit card with an average of four each, while in Europe the number of cards carried per inhabitant
ranges from 0.8 to 3.9.
Furthermore, as stated by the Word Payments Report 2019, non-cash payments are expected to reach the number of more than 1 trillion transactions by
2022: that's a huge quantity!
At the same time, the downside that comes from using credit cards is the presence of fraudulent transactions: in 2018, $24.26 Billion was lost due to payment
card fraud worldwide and that loss has been steadily increasing for the last
five years. As a consequence, different techniques were born to prevent fraudulent transactions from taking place and especially to detect them: here we will
address the anomaly detection process applied on the Credit Card Fraud
Detection dataset.
- Dataset analysis
- Dealing with imbalanced dataset
- Undersampling
- Oversampling
- Combining undersampling and oversampling
- SMOTE
- Anomaly detection
- Hyperparameters optimization
- Classification
- Logistic Regression
- K-Nearest Neighbors
- Support Vector Classifier
- Decision Trees
- Random Forest