Automated Clustering for Data Analytics

Today, auditors must consider the risks of material misstatement due to fraud during the financial statement audit (Messier, Glover, and Prawitt 2016). Current audit guidance recommends the use of data mining methods such as clustering to improve the likelihood of discovering irregularities during fraud risk assessment (ASB 2012). Unfortunately, significant challenges exist relative to using clustering in practice, including data preprocessing, model construction, model selection, and outlier detection. The traditional auditor is not trained to effectively address these complexities. One solution entails automation of clustering, thus eliminating the difficult, manual decision points within the clustering process. This would allow practitioners to focus on problem investigation and resolution, rather than being burdened with the technical aspects of clustering. In this paper, automated clustering is explored. In the process, each manual decision point is addressed, and a suitable automated solution is developed. Upon conclusion, a clustering application is formulated and demonstrated.