This project focuses on exploring different machine learning methods using Kaggle's Credit Card Fraud.
This project focuses on exploring machine learning methods through their application on Kaggle’s credit card fraud detection dataset available here: (https://www.kaggle.com/dalpozz/creditcardfraud). Labels are given in the dataset making this is essentially a classification problem with a variety of approaches one can take. The links and references given in the Reference section point to a number of papers that talk about different approaches used on this dataset and related credit card fraud datasets; these will hopefully serve to compare our performances with the ones achieved by those authors.
After exploring the data through visualizations and summary statistics, we will implement:
- Random forest methods
- Logistic Regression
- Support Vector Machine
- a Neural Network using hyperbolic tangent as its activation function.
These methods will be implemented in Python with the use of sklearn, scikit, numpy and pandas libraries. We plan on building several models based on the inclusion and exclusion of significant predictors and compare their accuracies per method to the ones reported in literature.