Feature Extraction of Fraudulent Financial Reporting Through Unsupervised Neural Networks

Abstract

This tutorial is based on modification of the professor nomination lecture presented two years ago in front of the Scientific Council of the Czech Technical University in Prague [16].It is devoted to the techniques for the models developing suitable for processes forecasting in complex systems. Because of the high sensitivity of the processes to the initial conditions and, consequently, due to our limited possibilities to forecast the processes for the long-term horizon, the attention is focused on the techniques leading to practical applications of the short term prediction models. The aim of this tutorial paper is to bring attention to possible difficulties which designers of the predicting models and their users meet and which have to be solved during the prediction model developing, validation, testing, and applications. The presented overview is not complete, it only reflects the author's experience with developing of the prediction models for practical tasks solving in banking, meteorology, air pollution and energy sector. The paper is completed by an example of the global solar radiation prediction which forms an important input for the electrical energy production forecast from renewable sources. The global solar radiation forecasting is based on numerical weather prediction models. The time-lagged ensemble technique for uncertainty quantification is demonstrated on a simple example.

Big data techniques in auditing research and practice: Current trends and future opportunities

“…Other approaches use supervised neural networks (Green & Choi, 1997;Krambia-Kapardis, Christodoulou, & Agathocleous, 2010) or unsupervised neural networks based on a growing hierarchical self-organizing map (e.g., Huang, Tsaih, and Lin (2014); Huang, Tsaih, and Yu (2014)) to build a financial fraud detection model. The approach proposed by Huang, Tsaih, and Lin (2014) involves three stages: first, selecting statistically significant variables; second, clustering into small sub-groups based on the significant variables; and third, using principal component analysis to reveal the key features of each sub-group. Huang, Tsaih, and Yu (2014) apply this model to 144 listed firms and find that the approach can effectively detect fraudulent activity.…”

A Survey of Automated Financial Statement Fraud Detection with Relevance to the South African Context

“…To our knowledge, unsupervised learning approaches have only been explored in Y.-J. Chen (2015), Deng and Mei (2009), Huang, Tsaih, and Lin (2014), Huang, Tsaih, and Yu (2014), Tsaih et al (2009). A thorough comparison of unsupervised learning methods has not been conducted in the FSF literature.…”

“…Thus, an unsupervised learning approach, which does not require a labelled data set, would more appropriate in the South African context. The unsupervised approach has been used on companies listed on the Taiwan and the Chinese stock exchanges and provided promising results (Deng & Mei, 2009;Huang, Tsaih, & Lin, 2014;Tsaih et al, 2009). As these are emerging markets, using an unsupervised approach could potentially provide good results when applied to a South African data set as South Africa is also an emerging market.…”