Master Defense – Interactive Trace Clustering

On March 10, 2020, Thaís R. Neubauer defended her master’s thesis. Prof. Dr. Sarajane Marques Peres supervised her and Prof. Dr. Marcelo Fantinato was the co-supervisor. Thaís worked with interactive clustering applied to business process mining.

In her thesis, she presented use cases for the constrained clustering algorithm Cop-K-Means, discussing its effects on trace clustering context, considering synthetic and real-world event logs. In the synthetic logs, Thaís explored the limitations of using Cop-K-Means in trace clustering. From this study, Thaís proposed new concepts that were important to systematize the analysis of such limitations. These concepts were called “space” and “force”. Thais intends to put these ideas for analysis by the international scientific community soon. Real-world logs used in this study refer to an incident management support system. In this context, Thais addressed the completion time prediction problem. The event logs were pre-processed using trace clustering and using what Thaís called as “interactive trace clustering”. According to the experiments carried out in this study, it was possible to achieve some improvement in the completion time prediction’s accuracy. Thaís’ work also brought another contribution: a systematic way for implementing the inclusion of human experts in the interactive trace clustering process. You can access the slides used in the defense session here (in Portuguese).

The evaluating board of this work had the following researcher:

  • Prof. Dr. Sarajane Marques Peres (chair) – USP
  • Prof. Dr. Sylvio Barbon Júnior – UEL
  • Prof. Dr. Ronaldo Cristiano Prati – UFABC

Within this work, Thaís and her supervisor are grateful for the direct contribution of former students Claudio A. L. do Amaral and Alexandre G. L. Fernandes, and for the specific guidance from the researchers Prof.dr.ir. Hajo A. Reijers and Prof. Dr. Xixi Lu from the University of Utrecht.

This work was supported by:

Related posts