Machine Learning

Algorithms .. Credit Card Fraud

Machine Learning

The machine learning use case in Pentaho Data Integration addresses credit card fraud detection, a critical business challenge where organizations typically lose about 5% of their yearly revenue to fraudulent activities*.

This hands-on lab demonstrates three powerful approaches to fraud detection:

First, it teaches how to use H2O (AutoML) to automatically discover well-performing machine learning models without extensive manual tuning.

Second, it covers supervised learning algorithms that can detect fraudulent behavior by learning patterns from historical fraud cases.

Third, it explores unsupervised learning methods to identify new, previously unseen types of fraud activities.

A key focus of this use case is handling imbalanced datasets, which is essential since fraudulent transactions are rare compared to legitimate ones. The use case provides practical techniques for properly classifying data when dealing with this common real-world challenge, ensuring that machine learning models can effectively identify fraud without being overwhelmed by the volume of normal transactions.

*Association of Certified Fraud Examiners (ACFE) "Occupational Fraud 2024: A Report to the Nations," published in March 2024

Lab Overview