Purchasing a house is one of the most expensive and long-term investments that most people will make in their lifetime. The ability to accurately estimate property values is of importance to homebuyers, homeowners, lending companies, insurance companies, tax assessors, and policymakers. In recent years, websites, such as Zillow, Trulia, and Realtor, have popularized the use of Automated Valuation Models (AVMs). AVMs utilize the characteristics of a property to provide a quick and relatively inexpensive estimate of the property's value. Unfortunately, the data used to train AVMs are often incomplete and filled with inaccuracies. We present a method employing a Dirichlet process for clustering properties in the presence of outliers.