KNIME for Finance - Fraud Detection: DBSCAN

Credit card Fraud Detection is a continuous challenge, requiring the identification of evolving fraud patterns. Due to the rarity and outdated nature of available fraud examples, adapting to new fraud patterns is difficult.

For this part on our series in Fraud Detection, we will be using another type of outlier-based classification approach called DBSCAN. DBSCAN is a popular clustering algorithm. Since it’s an unsupervised algorithm, a labelled dataset is not necessarily needed, however, it is useful to have such a dataset to rate the performance of the model.