There has been a significant increase in the amount and accuracy of mineral data (from resources like Mindat, MED or the GEMI) and the improvements in technological resources make it possible to explore and answer large, outstanding scientific questions, such as, understanding the mineral assemblages on Earth and how they compare to assemblages and localities on other planets. In the last couple of years, affinity analysis methods have been used to:1) Predict unreported minerals at an existing locality, 2) Predict localities for a set of known minerals[1]. We’ve chosen to call this application “Mineral Association Analysis”[2]. Affinity analysis is an unsupervised machine learning method that uses mined association rules to find interesting patterns in the data. Most of the metrics used to evaluate market basket analysis methods focus on either the ability of the model to ingest large amounts of data[3], or using a metric based comparison of various algorithms used for association rule mining[4], or on evaluating the rules mined to more efficiently generate association rules[5]. However, when patterns generated in an unsupervised method are used to predict the occurrences of entities such as minerals, there needs to be a way to evaluate the predictions made by the model. It’s in such an area that there has been very little work. In this abstract, we explore the development of a new method to evaluate the results of association rule mining algorithms specifically when used when the association rules generated are utilized in a predictive setting. [1] Prabhu et. al (2019). In AGU Fall Meeting Abstracts (EP23D-2286). [2] Morrison et al. Nat. Geo. (2021) In Prep. [3] Agrawal et al. (1993) SIGMOD’93. [4] Sharma et al. (2012) IJERT 1(06). [5] Üstündağ and Bal (2014) Proc. in Comp.