PAIS Lab seminar: Support of Decision-Making in IT-project Risk Assessment under Z-numbers / Discovering uml diagramm from event logs using process mining
At the session of our seminar, Ekaterina Goleva (Master's student at the HSE CS Faculty) delivered a report "Support of Decision-Making in IT-project Risk Assessment under Z-numbers" and Maxim Savitsky (Master's student at the HSE CS Faculty) delivered a perort "Discovering uml diagramm from event logs using process mining".
Abstract «Support of Decision-Making in IT-project Risk Assessment under Z-numbers»
The quality of provided by the analysis decision is directly related to the information given for the decision-making. However, usually the information is uncertain, imprecise and/or incomplete. Expert knowledge and qualified opinions are using in the area of the analyzed problem to decrease the uncertainty of a final solution. On the step when the expert opinions were received appears another problem – how to convert expert opinions expressed by the phrases and sentences of natural language in numeric form to make the analysis? Expert opinions formalized to some degree through numerical values can be called expert judgments.
Classical fuzzy set theory can be applied for converting expert opinions in the expert judgements, but some degree of certainty can be lost during the converting of expert opinion. L.A. Zadeh introduced the term ‘Z-number’ (definition will be given further in text) in 2011. In comparison with the classical fuzzy set theory a Z-number adding the parameter for reliability of expert concerning his/her judgment to the information given by expert about the problem, it allows to extract more input data for analysis and increase the accuracy of the results. However, modern methods of analysis via Z-numbers are still not mature. Existing methods are applying classical fuzzy set theory for converting Z-numbers into fuzzy numbers to provide calculations required by analysis (defuzzification of Z-number). The disadvantage of defuzzification is the weights approximation (expert judgement and its reliability) that leads to the loss of analysis accuracy. The approach based on Z numbers is still a new direction in uncertain computations because related works are coming back to the methods offered by the classical fuzzy set theory that converting two parameters into single generalized value instead of using for computation both of Z-number parameters as independent values (value of expert judgment and expert reliability).
Abstract «Discovering uml diagramm from event logs using process mining»
The work is dedicated to the graphic representation of the event logs. The key role is given to a unified view of date - UML, in particular the sequence diagram. In the analysis of event logs we shared components on the basis of some predifined rules. On example of a few logs of real systems the are searching for applicability of those or other partitions. Algorithm of particion is a part of a work.