Keskeiset käsitteet
Integrating a Naive Bayes classifier with sampling improves representativeness and riskiness balance in audit evidence selection.
Tiivistelmä
The content discusses the integration of a Naive Bayes classifier with sampling techniques for auditing purposes. It explores the challenges faced by auditors in processing excessive data and drawing audit evidence. The study focuses on classifying data using machine learning to avoid bias, maintain randomness, and target riskier samples. Three approaches are discussed: user-based, item-based, and hybrid, each aiming to draw representative audit evidence. Experiments demonstrate the benefits of unbiased sampling, handling complex patterns, correlations, and unstructured data efficiently. Limitations include classification accuracy output by machine learning algorithms and threshold variations affecting sampling outcomes.
Directory:
Introduction
Challenges faced by auditors in processing excessive data.
Literature Review
Studies integrating machine learning with sampling.
Naive Bayes Classifier
Application of the classifier for selecting audit evidence.
Results
Three experiments demonstrating benefits and limitations of machine learning integration.
Discussion
Benefits and limitations of integrating a Naive Bayes classifier with sampling.
Tilastot
"Three experiments show that sampling using machine learning integration has the benefits of drawing unbiased samples."
"Calculating the AUC from Figure 4 obtains 0.965 (Equations (3)-(4)), 0.953 (Random forest classifier), and 0.955 (Support vector machines model with a radial basis function kernel)."
"Table 2 lists other metrics for demonstrating classification accuracy on this confusion matrix."
Lainaukset
"Auditors may hybridize those user-based and item-based approaches to balance representativeness and riskiness in selecting audit evidence."
"Sampling using a Naive Bayes classifier has limitations."