Kenneth O. Cogger

Neural Network Detection of Management Fraud Using Published Financial Data

This paper uses Artificial Neural Networks to develop a model for detecting management fraud. Although similar to the more widely investigated area of bankruptcy prediction, research has been minimal. To increase the body of knowledge on this subject, we offer an in-depth examination of important publicly available predictors of fraudulent financial statements. We test the value of these suggested variables for detection of fraudulent financial statements within a matched pairs sample. We use a self organizing Artificial Neural Network (ANN) AutoNet in conjunction with standard statistical tools to investigate the usefulness of these publicly available predictors. Our study results in a model with a high probability of detecting fraudulent financial statements on one sample. The study reinforces the validity and efficiency of AutoNet as a research tool and provides additional empirical evidence regarding the merits of suggested red flags for fraudulent financial statements.

Detection of management fraud: a neural network approach

The detection of management fraud is an important issue facing the auditing profession. A major contributor to this issue is the Loebbecke and Willingham (1989) conceptual model for the detection of management fraud. A cascaded Logit approach using the Loebbecke and Willingham model was developed (Bell et al., 1993). The present study offers an alternative approach using artificial neural networks (ANNs). This paper develops a successful discriminator of management fraud using both the generalized adaptive neural network architectures (GANNA) and the adaptive logic network (ALN) approaches to designing neural networks. The discriminant functions can distinguish between fraudulent and non-fraudulent companies with superior accuracy to the cascaded Logit results.<<ETX>>

Proposals for research in time series forecasting

Abstract This paper reviews to developing lines of forecasting research which have received limited attention in the past and which appear to promise significant improvements in modeling ability and forecasting accuracy. Previous work is summarized, some new results are presented, and suggestions are given for fruitful new avenues of investigation.

Rounding errors in autoregressive processes

Abstract Time series observations are often rounded, but are modelled as though they were continuous and no rounding had occurred. This paper examines the impact of rounding on the estimation of parameters in autoregressive time series models, deriving appropriate adjustments for the estimates of the true parameters when using rounded data. Analytical results are presented for the asymptotic case, and simulation results are reported for the case of moderate sample sizes. The adjustments are simple and easily implemented, and do not require additional parameter estimation beyond the usual maximum likelihood analysis. Based on the findings of our analysis, we offer specific recommendations on how to adjust the parameter estimates in practice for different levels of rounding.

Piecewise Linear L1 Modeling

Lester Taylor (Taylor and Houthakker 2010) instilled a deep respect for estimating parameters in statistical models by minimizing the sum of absolute errors (the L1 criterion) as an important alternative to minimizing the sum of the squared errors (the ordinary least squares or OLS criterion).

A Comparative Study on Eigen Weight Vectors

There are infinitely many eigen weight vectors which can be constructed for any given data of estimated weight ratios. Assuming that the reliability of these individual ratios can be different, we study the properties of the different eigen weight vectors, including that of Saaty and that recently proposed by Cogger and Yu. A general framework for the construction of eigen weight vectors based on the reliability of the data will be proposed and discussed.