This repo consists of a python notebook that demonstrates how to predict fraud transactions from a highly imbalanced dataset. The data is from a Kaggle dataset 'Credit Card Fraud Detection'. The data states that the features are scaled and anonymized for security reasons. The techniques used in the notebook for dealing with imbalance are:
- Undersampling
- Oversampling through SMOTE
The algorithms used are:
- LogisiticRegression
- KNearest
- Support Vector Classifier
- DecisionTreeClassifier
- Deep Neural Network through Keras.