| Summary: | 碩士 === 華梵大學 === 工業工程與經營資訊學系碩士班 === 99 === With the advance in technology, companies have applied information systems to deal with the complicated workflow activities. These activities are recorded by information systems in the form of so-called “workflow log”. Recently, various process mining methods were proposed to extract the process model from workflow log. To correctly explain the behavior of workflow log from the mined process model, the conformance of workflow log and the mined process model has become an important research issue.
Currently, the fitness (f) is the main measure adopted by process mining algorithms for conformance checking. Fitness only forcuses whether the mined process model can simulate the behavior of workflow log, but ignore the extra behavior generated by the mined process model but not appeared in the workflow log. This will result in the misunderstanding and error decision about the analyzed business process for managers. Therefore, another measure, precision (a’B), was proposed by researchers for conformance checking of process mining.
As described above, the existing process mining methods may derive a process model with high fitness and low precision from workflow log. This research has applied a two-staged clustering method (AHC Ward’s+K-Means) to improve the precision of process mining. Besides, the structure of workflow log resulting in the low precision is also analyzed.
|
|---|
