This blog post examines how process mining identifies inefficiencies in business operations through event log analysis and leads to tangible process improvements through the modification and extension of process models.
A business process refers to the entire sequence of tasks an enterprise or organization performs to create customer value. The procedure of diagramming this task sequence as a workflow is called process modeling, and the resulting diagram is called a process model. Process models consist of task activities and the paths between them. Based on this, it is necessary to verify, analyze, modify, supplement, and improve the process model to ensure it operates efficiently. Process mining is a technique utilized within these improvement efforts. Process mining bridges the gap between simulation-like analysis techniques centered on pre-defined process models (before actual event logs are collected) and data-centric analysis techniques like data mining that do not consider the process.
Process mining is the task of extracting valuable information related to the process from event logs obtained through information systems. Event logs refer to the records of business process execution accumulated within information systems, serving as the starting point for process mining. Event logs are represented as two-dimensional tables composed of rows and columns. Events generated by business activities are added as rows, and each column records the attributes of the corresponding event. The essential attributes recorded are the case ID, activity name, and occurrence time; additional attributes can be included for various analyses. Event logs are merely raw data providing information directly useful to users, necessitating a process to transform them into analyzable information. Process mining encompasses three types: process discovery, compliance verification, and process improvement.
Process discovery refers to the task where a process analyst uses algorithms to derive a process model from event logs; the analyst can perform this task without requiring specific domain knowledge. If the derived process model is overly complex, making meaningful analysis difficult, techniques like fuzzy mining or clustering can be applied. Fuzzy mining simplifies the process model by removing or merging activities with low execution frequency and eliminating paths between activities. During this process, thresholds can be set for activities and paths appearing in the process model to control its complexity. Clustering is a technique that groups similar cases together. When deriving a process from the entire event log results in a complex process model, clustering can be applied to partition the event log into multiple segments. Applying process discovery techniques to these subdivided event logs reduces the complexity of the process model.
Fitness verification is the procedure of comparing the results derived from the existing process model with those from event log analysis to determine the degree of agreement between the two sets of results. In this process, discrepancies may arise between the existing process model and the results derived from the event logs. For example, even if the existing process model is appropriate, there may be cases where the personnel responsible for the task do not adhere to it. In such cases, the actual work practices occurring in the real world must be corrected.
Conversely, if the results from event log analysis are deemed more appropriate, the existing process model must be modified. Process enhancement is divided into two types: “modifying” the existing process model, and “extending” the discovered process model by adding supplementary information obtained from event log analysis, such as task execution times or personnel.
An example of extension involves visualizing bottlenecks and rework flows within the process model derived from event logs. This enables clearer identification of problematic areas within the process.
Process mining, grounded in data science, enables process analysts to collaborate with business experts to diagnose issues in a company’s business processes and derive improvement strategies. With the recent advancement of information systems and data processing technologies, the scope of process mining applications continues to expand. It is increasingly utilized as a critical tool for enhancing operational efficiency and decision-making precision within enterprises.
