Opinion detection of public sector financial statements using K-nearest neighbors

The identification of ethical violations committed by the auditor is very difficult to do. Artificial intelligence offers anomaly detection as an alternative method for detecting the opinion anomaly which can be an early indicator of the opinion trading occurrence. This paper proposes the use of original features from public sector rather than the use of modified features from the private sector to be applied in opinion detection in public sector. By using 60% Holdout validation, 1-NN classification showed that original featured from the public sector outperformed the modified featured from the private sector by 5.82% through 13.10% under F-Measure Criterion and by 4.22% through 9.56% under AUC criterion.