New tool available
SCCH Process ANT
In a typical (economical) process analysis the processes are identified by observing the actions, random samples or by conducting interviews with the employees. The results are gathered with a lot of effort and, nevertheless, represent in many cases more the de jure (should) than the de facto (is) process (especially with respect to rare misbehavior) and there are no concrete figures regarding the possible execution paths, frequencies and times.
The aim of the technical process analysis (process mining) is to detect, monitor and improve real processes via extraction of knowledge from event logs. Events thereby refer to a multiplicity of actions in a company. Depending on which application is examined, the events can refer e.g. to executing a financial transaction, receiving an electronic product request, screwing metal parts or calling a method in a software component.
The resulting process model is called de facto (or is) process model, since it pictures the real process execution. This is also the main difference between process mining and classical business process management (BPM). BPM is based on a top-down approach, i.e. a process model is defined (also called de jure or should process model) and then this process is implemented by IT and executed in the production process. However, deviations may occur in practical execution, thus compliance cannot be guaranteed. In contrast, process mining provides a bottom-up approach. It is based on the executed production process, identifies the corresponding data and creates a fitting process model (de facto process model) with detailed information about execution paths and frequencies. In an optimal case, both, the de jure and de facto process model, are available, so that discrepancies can be identified by a conformance check.
Mapping of columns to the analysis fields with fields for textual (categorical c.) and numerical (numerical c.) meta data.
Generated graph is displayed in the program and stored as a .graphml file
The data, which is used for process analysis, generally already exists in the company, e.g., as event logs of machines or collected and processed in data bases. These data are analysed and connected with each other. The resulting de facto process model allows to discover bottlenecks, identify outliers and misbehavior, recognize patterns and to prevent process abortion. A detailed performance analysis will further identify optimization potentials and improve, e.g., execution order and execution-/waiting-times. Further potentials of process mining are process monitoring (regular analysis) as well as process simulation and prediction. The resulting knowledge about the actual course of action in a company serves as basis for the implementation of software systems, performance analyses and process improvements, checking adherence to processes, as well as process monitoring and simulation.
To support a technical process analysis, we developed the tool SCCH Process AnT at SCCH. Data can be imported from CSV files and from databases. After selecting a process and mapping the columns to the corresponding analysis fields, data can be imported and the desired period selected. A filtering of outliers (noise) is possible by setting a threshold value. The resulting process graph is displayed in the graphical user interface and saved as a .graphml file so that it can also be opened by other tools (e.g., yEd).
Characteristics of SCCH Process AnT are:
- High-performance process analysis for large data amounts
- Representation of the entire process model (100%) and filtering of outliers
- Detailed analysis and presentation of key performance indicators (absolute and relative frequencies), time values (execution and waiting times) and metadata (for example, costs, resources or quality indicators)
- Comparision of de jure (should) and de facto (is) process models
In the process graph next to every node, the execution time (mean and standard deviation) as well as an assessment of possible meta data is displayed. Every edge is marked with the absolute and relative frequency as well as waiting time (mean and standard deviation).Optimization potentials are identified based on process paths, time values and meta data. Considered should be infrequent paths, which may indicate outliers and misbehavior, as well as paths connecting a process step with the artificial end-event (may represent process abortions). Furthermore, considering time values, long waiting times are an evidence for bottlenecks and execution times with a high standard deviation indicate improvement potential.
For further information or a demonstration of SCCH Process AnT please contact Dr. Christine Natschläger.