This document discusses fraud detection techniques using data mining. It describes how data mining and statistics can help anticipate and quickly detect fraud. Three main techniques are discussed: Bayesian networks, decision trees, and backpropagation. Bayesian networks construct models to describe fraudulent versus legitimate behavior. Decision trees induce a tree to generate classification rules. Examples of rules generated from a decision tree for fraud detection are provided. The document also briefly mentions existing fraud detection systems such as a fuzzy logic system and self-organizing feature map.