| Summary: | 碩士 === 國立臺灣科技大學 === 工業管理系 === 98 === Recently, workflow automation has been widely applied in industry. Log files stored the activities sequence for each case can be used to construct the process model in terms of the developed process mining algorithm. However, due to the complexity of the process, the mined model might be very complicated and hence difficult to view and to analysis. Cluster the stored cases and to mine each group of cases can simplified this issue.
Currently, the developed workflow clustering algorithm tends to compute the sequential relationship of the activities based on the whole record of the case. The computational efficiency will decrease a lot for a case with long sequence of activities.
In this research, an approach based on blocking the log into several group and clustering the activity data in each group will be proposed. This approach ignores the sections of the records having the common sequential relationship and addresses on the portions where cases have diverse string of activities. That is, based on the mined model, the sequence of activities in the log will be classified as several blocks. Then the activities in each block will be clustered and the inter-block activities relationship also will be analyzed. By applying this approach, the computation efficiency will be increased for the log with long string of activities but contained common sequential relationship. In addition, the attributes for each group of cases can be analyzed by identifying the features of individual block of activities.
|
|---|
