Analisa Kualitas Auditor Di Kantor Jasa Akuntansi Menggunakan Multi Layer Perceptron

Abstract

Auditing practices are generally carried out by auditors who join a partner organization consisting of senior auditors and junior auditors. The quality of this audit is generally as-sessed from the ability of the auditor to find errors in the auditee's financial statements and report their findings. This study compares the results of the classification of auditor quality at the Accounting Services Office (KJA) Azhar Maksum & Partners between the multi-layer perceptron classification model that uses the ReLu, Tanh and Sigmoid activa-tion functions. Evaluation was measured using k-fold cross validation with variations in K values of 5 and 10 to measure accuracy, F1, precision and recall. The results obtained from the average cross validation value are that the ReLu activation function is the best activation function to be used in the quality classification model of KJA Azhar Maksum & Partners auditors, it can be seen from the average accuracy value of 82.7%, the average F1 value of 83.4%, the average precision value is 88.9% and the average recall value is 82.7%. Also, the Sigmoid activation function is the worst activation function to be used in the quality classification model of KJA Azhar Maksum & Partners auditors, it can be seen from the average accuracy value of 69.2%, the average F1 value of 56.6%, the average precision value of 47.9 % and the average recall value is 69.2%.