Skip to content

Gianluca124/Knime_Classification_Credit-Card-Fraud-Detection

 
 

Repository files navigation

Credit card fraud detection

Project for Machine Learning course 2021/22

In the financial field, one of the most important aspects is the detection of fraudulent transactions on credit cards. This may be "authorized" fraud, where the cardholder processes a payment to a suspicious account, or "unauthorized" fraud, where the cardholder does not provide any authorization for the transaction, which is carried out by a third party. From an economic point of view, it is essential for financial institutions to be able to detect these transactions in advance: in 2018 the UK suffered credit card fraud amounting to £844.8 million. Nowadays, credit card transactions are very safe, and this is also due to the great effort made by financial institutions in the field of "cyber crimes" detection. In order to complete this task effectively, one of the objectives to be achieved is to identify the best machine learning technique, and in particular the best classifier, that is able to recognize, and consequently block, these types of transactions. The project focuses on counteracting the negative effects that a dataset with unbalanced classes, as in this case, has on finding and estimating the best classifier for forecasting fraudulent transactions. The performance of the models will be evaluated not only through the usual metrics, but also taking into account the computation costs and the quality of the anthropomorphic services provided to the cardholder.

Workflow preview

English version of the report incoming!

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%