Business process mining techniques use event logs recorded by information systems to extract and discover useful process and organizational information. Over the past two decades, many new and powerful process mining techniques have been developed by researchers and software vendors. Although numerous case studies demonstrating the applicability of process mining have appeared in the literature, there is still no application of process mining in the financial sector.
In this paper, we investigate the applicability of process mining to an investment and securities firm, specifically investigating its management accounting processes. Using a set of process mining techniques available in the ProM framework, we examine and discover clear differences between the AS-IS model provided by the financial company and the process model obtained from the event logs.
Introduction
In short, the domain of process mining research can be categorized into its (1) techniques and (2) applications. This study aims to add to the growing process mining application literature with an application in the financial domain, which tends to contain unstructured, non-linear, complex processes. This study therefore addresses the possible need for a different approach needed to tackle applications of process mining in the context of less than adequate case sizes.
Related Work
Business Process Reengineering
Managerial Accounting
Process Mining
Traditionally, process mining research and literature revolves around discovery, where information about the original process model, organizational context, and execution properties is extracted from execution logs (Mans et. al, 2008). In the discovery phase of process mining, process mining techniques such as α-algorithm mining and heuristic mining are used to produce visual representations (i.e., Petri nets) of real processes obtained from event logs. In other words, process discovery algorithms automatically construct process models based on the behavior observed in the event log (Rozinat et. al., 2010).
However, the discovery is not just limited to the extraction of control flow and process models; recent process mining techniques are becoming increasingly popular. This discovery phase plays an important role in laying the foundation for the analyzes of real processes. In the conformance phase of process mining, the process models extracted from reality (i.e. event logs) are compared with an AS-IS model.
Conformance checking can be used to detect deviations, locate and interpret these deviations, and measure the severity of these deviations (van der Aalst & Günther, 2007). Consistency checking techniques assess the relationship between process models and the reality represented in event logs through orthogonal dimensions of consistency (i.e. relevance, accuracy, generalizability, and structure) (Adriansyah et. al., 2010). In the extension phase of process mining, the process model is extended with new aspects or perspectives to enrich the model with data in the event log (Mans et. al., 2008).
Some examples of extension are illustrating the bottlenecks in a process model by analyzing the event log or using decision point analysis to detect data dependencies that affect the routing of a case (i.e. how data affects the choices made in the process based on previous process executions) ( Sang & van der Aalst, 2008).
Research framework
Case Analysis
Log Description
According to the AS-IS model, business management handles core core tasks and is intertwined with all other departments; other departments perform smaller tasks and interact and provide relevant information to assist the business management department in evaluating the overall accounting (ie expenses, profits, wages, etc.) of the firm. Less vital tasks such as payroll expenses, departmental and product profit and loss calculations, and other sales and administrative overhead calculations are handled by these support departments in this process. From the perspective of managerial accounting, such departments play a supporting role for the business management department, where the main tasks are performed.
A more specific analysis of the handover of work will be discussed later using social network analysis. From Figure 4 alone, it can be seen that the monthly profit calculation process is a simple and straightforward process. Using process mining techniques, we examine whether the AS-IS model is an accurate representation of the actual behavior of the process recorded in the event log.
Mining
- Control Flow Perspective
- Conformance Analysis
- Performance Perspective
- Organizational and Social Network Perspective
To better visually represent the differences between the AS-IS model and the model obtained from the heuristic mining algorithm, Figure 8 shows the overlay of the workflow extracted from the event log onto the AS-IS model. The dashed lines with arrows represent the original AS IS workflow paths that were not found to match the event log workflow paths. This can be interpreted as only 40.2% of the event log data "fit" the AS IS model.
According to the extracted α-mining and heuristic algorithm model (Figures 6 and 7), the initial and final tasks of the process are very ambiguous, almost undefined. The creation of new models using the control flow mining techniques was repeated after redefining the start and end tasks in the event log. Conformance analysis of heuristic mining algorithm with artificial start and end tasks. Table 2 contains the summary of the conformance fitness metrics assessed in this section.
It can be observed that the fitness of the AS-IS model with loops and the α-mining model are quite similar. However, the fit of the heuristic models is significantly better than the AS-IS and α-mining models. The heuristic model is a much better representation of the actual process than what is recorded in the event log.
As a side note, bottlenecks 1 and 11 can be omitted due to the additions of the artificial start and end tasks in the event log. Our final analyzes concern the organizational structure and social network aspects of the company, using the event log. For example, the third group of the Business Management Department specializes in tasks relating to 'Other Sales and General Administrative Expenses'.
Discussion
Human Resources, Payment Services, General Affairs, Training and E-Business Departments), who are responsible for less vital tasks (salary expenses, sector and product profit and loss calculations, as well as other calculations of sales and general administrative expenses) of The process of calculating the monthly profit and loss was carried out earlier in the process, then followed by the most important tasks which were carried out by the Business Management Department. Not only is this compliance evidence of the work allocations that are distributed among the support departments, then their results are collected for the main important tasks to be performed by the Business Management Department, but it also provides a closer insight into how the process happens. A major advantage of using process mining as a complementary tool in BPR is that performance-related problems within the process can be visually identified easily, which was previously difficult to find with its traditional methods. process evaluation.
Through performance analysis, bottleneck specifics can be identified based on which task is problematic, who is performing the task, and how severe the problem is in itself and the process as a whole. A complete and workable solution to fixing such obstacles is difficult to create with current process mining techniques alone. Regarding the organizational and social network perspective of our research, in our opinion, there were no outstanding details that were worthy of special consideration or attention.
To gain more insight into the organization's social network, it is recommended to conduct analyzes at the individual or group/team level. Overall, the heuristic algorithm model shows that process mining can provide insight into how processes are actually executed. Furthermore, because process mining uses concrete data from information systems logs, it is essentially immune to the problem of misleading or missing information when assessing or diagnosing the status of an organization through interviews and surveys.
Another goal of our research was to highlight some of the strengths of process mining and how process mining can be a complementary tool to improve the current methodologies of BPR, thus hoping to ultimately and indirectly increase its success rates.
Conclusion
2008) have already argued for the necessity of a process mining evaluation framework that allows a) process mining researchers to compare the performance of their algorithms and b) end users to assess the validity of their process mining results. This lack of evaluation techniques motivates the development of such evaluation methods or frameworks for process mining results in future research, as well as to further strengthen current process mining tools and create more robust algorithms for future efforts. Since the event log data consisted of monthly P&L process records, each case represented process events within a month.
Although the certainty that the cases reflect the actual process can be increased by obtaining more cases, unlike the statistics domain where much research has been done in terms of calculating acceptable sample and data size, the process mining domain is still growing and in our opinion , badly needs research to specify how many cases are required to perform proper process mining analysis. Furthermore, with regard to case sizes, exceptions must be made for applications of process mining research to low case sizes because not all event log data have the abundance of cases. Unlike statistics, increasing sample (case) size tends to be more difficult if not impossible for event log data, and thus we also present the need for a framework for process mining research and applications for small case sizes.
Automated discovery of workflow models from hospital data', Proceedings of the ECAI Workshop on Knowledge Discovery and Spatial Data, p. Discovering Distributed Processes in Supply Chains', Proceedings of the International Conference on Advanced Production Management Systems (APMS 2002), p. 'Process Mining Applied to the Step Slice Test Process in ASML', IEEE Transactions on Systems, Man, and Cybernetics, Part C, 39(4), pp.
The need for a process mining evaluation framework in research and practice', BPM Workshop 2007, Lecture Notes in Computer Science, vol.
