Decision Making by a Novel Ensemble Mechanism

Fraudulent financial reports (FFR) have caused serious turbulence in financial markets and have damaged auditors' reputations. While auditors are the last line of defence in detecting FFR, many auditors lack the sufficient experience to deal with the related tasks. This investigation introduces hybridized model (HM) which incorporates four main parts: feature selection ensemble, extreme learning machine (ELM) ensemble, performance evaluation by multiple criteria decision making (MCDM), and knowledge generation, to alleviate auditing risks. The feature selection ensemble is grounded on ensemble learning. The advantage of this ensemble combination is its ability to catch errors made by an individual technique. ELM with superior generalization ability was adopted as the basic classifier for the HM. The study further decomposed the ELM's inherent structure to yield comprehensible rules in logical statement. Furthermore, the knowledge visualized process is supported by real example, can assist auditors who must allocate limited resources to make reliable decisions.