This document discusses using PageRank for anomaly detection in healthcare fraud. It summarizes an approach that computes similarity between healthcare providers based on procedure codes, builds a graph, runs personalized PageRank to score providers, and identifies anomaly candidates as those with high scores but in the wrong specialty. The Medicare Part B dataset is used and the implementation leverages Apache Pig on Hadoop to compute similarities and PageRank scores at scale. Examples are given of providers identified as anomalies, such as an internist billing specialty procedures or an otolaryngologist billing plastic surgery codes.