Problem transformation methods for prediction of opinion and exceptions in financial statements audit reports: Case for financial statements audit in central Kalimantan province

Abstract

The previous research related to financial statements audit mostly used single-label classification, such as opinion prediction, opinion identification, and opinion detection. We propose the use of multi-label classification to predict the “opinion and exceptions” using data from financial statements audit reports in Central Kalimantan province. We use financial ratios as attributes as well as opinion and exceptions as labels. In this research, we use three of Problem Transformation Methods, namely Binary Relevance (BR), Classifier Chains (CC) and Random k-labelsets (RAkEL), where each of will be combined with three of base classifiers such as J48, SMO, and Random Forest. The best evaluation metrics results for Hamming Loss is 0.19, for One-Error is 0.253, for Rank Loss is 0.16, and for Average Precision is 0.793.