Data Mining Applications in Accounting and Finance

Introduction

Data mining has been used in a variety of business applications, such as consumer buying pattern prediction and credit card default prediction, but recent research studies in accounting and finance have applied data mining techniques for classification and prediction of events such as firm bankruptcy and auditor changes. Data mining may be a useful tool for accounting and finance applications because of the huge volume of financial data that impacts business decisions. In this chapter, we summarize three published research studies in which we applied various data mining applications using accounting and other data for classification and prediction decisions, and we identify important issues to consider when applying current data mining tools.

The three research studies we summarize in this chapter are Kwak, Eldridge, Shi, and Kou (2009), Kwak, Eldridge, Shi, and Kou (2011), and Kwak, Shi, and Kou (2012). In these studies, we evaluate the relative usefulness of a variety of data mining approaches primarily by comparing overall accuracy rates for predicting material weaknesses in internal controls (Kwak et al., 2009), auditor changes (Kwak et al., 2011), and bankruptcy (Kwak et al., 2012). We compare the results of multiple criteria linear programming (MCLP) with those of linear discriminant analysis and Decision Tree based See5 (DT) in Kwak et al. (2009) and with those of classification and regression tree (CART), linear logistic regression (LLR), and Bayesian network in Kwak et al. (2011). The CART, LLR, and Bayesian network applications are three of the 13 data mining models in Witten and Frank’s (2005) WEKA Data Mining Workbench, and Kwak et al. (2012) apply all 13 data mining models.