Article: Classification of the Priority of Auditing XBRL Instance Documents with Fuzzy Support Vector Machines Algorithm.
Detecting fraudulent documents is time-consuming if the total amount of documents to be detected is large. Any technique, which can help to quicken the detection of fraudulent documents, is welcomed. Benford's law is one of such techniques.
Guang-Yih Sheu conducted an research on XBRL and the survey result published on the journal of autonomous intelligence.
This study develops a new application of fuzzy support vector machines algorithm. Fuzzy support vector machines models are constructed to separate more possibly fraudulent XBRL instance documents from others. The dependent variable in these fuzzy support vector machines models is a fuzzy variable describing the inconsistent conformity of an XBRL instance document to the Benford’s law. The independent variables are the price-to-book and equity ratios. Auditors may use the resulting fuzzy support vector machines models to determine which XBRL instance document is audited first. The theoretical background of determining which XBRL instance document is more possibly fraudulent is Benford’s law .
It has been used the proposed fuzzy support vector machines models to classify XBRL instance documents, which were presented by companies with full-cash delivery stocks. The goal is finding more possibly fraudulent XBRL instance documents. The misclassification rate is less than 30 %.
This study demonstrates that the machine learning technique (e.g. the fuzzy support vector machines algorithm) can improve the way conventional auditors work. It will denote the main evidence of applying a future project of training smart auditors funded by the Taiwan’s ministry of education.
For more information, please visit: