Page 36 - 臺大管理論叢第33卷第1期

auditors,
tasks
which
a
multi-objective
to
is
appropriate
equation
�
�
�
) .
+ ∑∑
(∑∑
+ ∑∑
�
�
�
��
�
�
��
�
�
��
�
�
� �
� �
� �
The results indicate that this method could prioritize important characteristics that
affected the performance of auditors; compared with the single-objective model, the
multi-objective model will first consider the high-performance audit portfolio, find out
the audit variables that improve the performance value through the overall model
analysis, and carry out the audit task according to the level of influence between the
audit variables—assignments to maximize conversion of audit performance values. For
example, auditors with high audit qualifications and high professional license teaching
An Integrated Data-Driven Methodology for Auditor Performance Appraisals and Auditor Assignment
Optimization
hours are assigned to audit tasks with low levels of complexity. Limiting the seniority
of three people in the audit team allows the team to have senior and junior auditors
in the audit team allows the team to have senior and junior auditors simultaneously
simultaneously (∑ ≤2, ∑ ≤2, ∑ ≤2). The existence of auditors
The existence of auditors helps promote senior
�
�� ��
�� ��
�
�� ��
�
personnel’s growth, leading to junior personnel’s development. Assignment methods
helps promote senior personnel's growth, leading to junior personnel's development.
show that audit experience, expertise, and task complexity affect audit performance and
Assignment methods show that audit experience, expertise, and task complexity affect
assignments, which is a finding that echoes current research findings (Asare and McDaniel,
1996; Tan, Ng, and Mak, 2002; Alissa, Capkun, Jeanjean, and Suca, 2014).
audit performance and assignments, which is a finding that echoes current research
Machine learning methods can improve the ability of audit-task scheduling model in
findings (Asare and McDaniel, 1996; Tan, Ng, and Mak, 2002; Alissa et al., 2014).
order to analyze critical factors and understand which combinations of audit specialties or
Machine learning methods can improve the ability of audit-task scheduling model
backgrounds perform best on tasks of varying complexity. When the number of auditors
and tasks solved increases to a certain number, the order of assignment of auditing tasks
in order to analyze critical factors and understand which combinations of audit
could be summarized as dispatching rules. Thus, in the future, audit supervisors will be
specialties or backgrounds perform best on tasks of varying complexity. When the
able to follow the audit dispatch rules for task assignment, and the encouragement of
learning in organization will also have a positive effect. For example, priority should be
number of auditors and tasks solved increases to a certain number, the order of
given to assigning auditors with professional training experience in their job qualifications
assignment of auditing tasks could be summarized as dispatching rules. Thus, in the
to the least complex tasks, and then assigning senior personnel with professional training
experience and financial background to the most complex tasks. for task
future, audit supervisors will be able to follow the audit dispatch rules
assignment, and the encouragement of learning in organization will also have a positive
4. Research Limitations/Implications
effect. For example, priority should be given to assigning auditors with professional
In terms of research limitations, although this study proposes a balance mechanism
training experience in their job qualifications to the least complex tasks, and then
based on further distribution of input data, the audit performance data has the
characteristics of time migration. Also, the audit work process may affect the subsequent
audit performance. Therefore, an ideal model should be able to process sequence data
and construct a complete audit analysis model. Follow-up can refer to the work shift
scheduling model, Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM),
and other related models of time series data (Ulmer, Thomas, Campbell, and Woyak, 2021;
Dahmen, Rekik, Soumis, and Desaulniers, 2020; Hochreiter and Schmidhuber, 1997), and
the audit data is further analyzed with a deep learning model, and long-term dependencies
in the sequence data are extracted to improve the prediction accuracy of the performance
evaluation model.
In terms of research variables, this study assesses auditors’ tasks based on difficulty
and number of days required , thus it lacks descriptions of the nature of auditing
28