Summary (797511)
Текст из файла
National Research University Higher School of EconomicsFaculty of Computer ScienceManuscriptAlexey Alexandrovich MitsyukStructure-Preserving Process Model Repair Basedon Event LogsPh.D. Dissertation Summaryfor the purpose of obtainingDoctor of Philosophy in Computer Science HSEMoscow — 2019This work was prepared at National Research University Higher School ofEconomics.Academic Supervisor:Prof. Dr. Irina A.
LomazovaProfessor, Dr.Sc., Faculty of Computer Science, NationalResearch University Higher School of EconomicsDissertation Committee: Prof. Dr. Alexander K. Petrenko (chairman),Professor, Dr.Sc.,Software Engineering Department, Institute forSystem Programming RASDr. Josep Carmona,Associate Professor, Ph.D.,Department of Computer Science, UniversitatPolitècnica de CatalunyaDr. Marco Montali,Associate Professor, Ph.D.,Faculty of Computer Science, Libera Università diBozen-BolzanoProf. Dr. Valery A. Sokolov,Professor, Dr.Sc.,Department of Theoretical Information Science,P.
G. Demidov Yaroslavl State UniversityProf. Dr. Alessandro Sperduti,Professor, Ph.D.,Dipartimento di Matematica, Università degli Studidi PadovaDissertation SubjectWe can not imagine our life without information systems. Processesin different domains (information technology, banking, healthcare, industrialproduction etc.) are supported by software systems that store and transform datarelated to these processes.
That is why the concept of process-aware informationsystems emerged in recent years [1, 2]. Systems of that type are based on coremodels describing process nature.Complex software and information systems can be designed using modeldriven software engineering approaches [3, 4]. They involve designing a systemstructure and processes using formal modelling notations, and then implementingcarefully elaborated blueprints in code.
However, the exact implementation of aprescriptive design model in real-life systems is a rare case. Moreover, processestend to evolve and erode during the system’s life-cycle. That is why a descriptionof a real system’s structure and behaviour usually differs from its design model.System owners want relevant, up-to-date models describing structure andbehaviour of their system according to real processes which are executed with thissystems’ support. This leads to the development of different methods to reverseengineer a system, analyse its structure and behaviour.In particular, the real behaviour of a software system can be studied byanalysing its event logs.
Process mining [5] is a field of research and technology thatproposes algorithms and methods for this kind of an analysis. One can discover themodel of a real system from event logs. Moreover, process engineers can diagnosethe discrepancies between observed (event logs) and modelled (process models)behaviours using conformance checking techniques.Conformance information can be used to improve or enhance models.
Forinstance, it is possible to repair process model using event logs [6], i.e. to constructa new process model which is based on a given initial model but conforms betterto a given event log. Using process model repair techniques a system owner maykeep a process model up-to-date. This thesis is devoted to a construction of suchalgorithms.Process model repair aims at improving the quality of a model1 withan additional constraint.
The repair should change as few model parts aspossible, thus preserving its structure. The latter differs model repair from processdiscovery, where the goal is to synthesize a model based on the given event logsuch that this model meets specified conformance characteristics. Thus, model1A model quality is calculated according to some quality criteria. In particular, a model needs to conform toa given event log.3repair is applied when existing process model is of value, and its owner does notwant to completely re-build it.The problem can be illustrated with the following example.
Figure 1shows a process model that does not fit2 the observed behaviour of a system.In particular, its fitness according to the event log is 0,97 (where 1 is perfectfitness).Figure 1: Given process modelWe may apply one of process discovery algorithms, and synthesize acompletely new model perfectly fitting the same log. For example, Figure 2 showsa model that has been discovered using Inductive miner with a zero noisethreshold from the event log.Figure 2: Model discovered from scratch using the Inductive minerThis model perfectly fits the event log. It models the same process withthe same set of activities, and contains transitions with labels from the same setas the initial model. However, these models are structurally different.
Note thatthe fitness of the initial model is almost perfect. Thus, inconsistencies in it arenot serious. Virtually, the model shown in Figure 1 can be repaired replacingtwo of its elements (transitions). Such problems do not always need the completere-discovery. Often, they can be repaired without substantial change in the model.In such a case, structure-preserving algorithms are relevant. This type ofmodel repair is difficult because it needs to find a balance between necessaryconformance to an event log, and desire to preserve a structure of the initial2This is one of many possible conformance metrics. A model perfectly fits an event log if it can replay allthe behaviour recorded in this log.
The fitness level shows what portion of log behaviour can be replayed by themodel.4model. The goal of this thesis is to propose a structure-preserving process modelrepair methods.The problem of this thesis is the model repair problem that can be definedas follows. The initial process model describes a behaviour of a system. Theevent log of this system does not conform3 to the process model according tosome predefined conformance criteria. In other words, this model does not fullyreflect the observed behaviour of the system.
The general model repair problemis to find a model (repaired process model ) that conforms the event log according to the mentioned above criteria, and is as similar to as possible.Besides, a repair procedure — if possible — does not change the process modelsignificantly.Main Results of this ThesisThe main research contributions of this thesis are as follows:I. A new modular approach (scheme) for process model repair based on eventlogs is presented. It is based on process model decomposition.
Process modelsare represented by workflow nets. The scheme may include various algorithmsfor model decomposition and sub-nets repair. Sufficient conditions of theapproach effectiveness are formulated.II. New algorithms are presented, performing local and non-local process modelrepair based on event logs. These algorithms implement the general modularscheme, and employ the divide and conquer principle.III. New algorithms for event log generation via process model simulation arepresented.IV.
Prototype implementations of the process model repair algorithms havebeen experimentally evaluated using event logs generated by the simulationalgorithms.Publication and PresentationThe results of this thesis have been published in international reviewedjournals and conference proceedings. For each paper we provide its status and theproportion of author contribution.3Note that in this thesis both following forms are used with the same meaning: a model conforms an eventlog and an event log conforms a model.5PublicationsMain results of this thesis are published in the following papers whichare split in three groups based on their formal status according to the rulesof HSE Dissertation Council in Computer Science4 . In particular, “First-tierpublications include papers indexed in the Web of Science (Q1 or Q2) or Scopus(Q1 or Q2) databases, as well as peer-reviewed collections of conferences thatappear in CORE rankings (ranks A and A*).
Second-tier publications are paperspublished in journals included in HSE’s list of high quality journals or indexedin the Web of Science (Q3 or Q4) or Scopus (Q3 or Q4) databases, as well aspeer-reviewed collections of conferences appearing in CORE rankings (rank B).”Third-tier publications — other papers.First-tier Publications1. Mitsyuk A. A., Shugurov I. S., Kalenkova A.
A., van der Aalst W. M.P. Generating Event Logs for High-Level Process Models //Simulation Modelling Practice and Theory. 2017. Vol. 74. P. 1–16. (SCOPUSand WoS-indexed Journal Paper, Q2; author contribution: 0,5) [7]Second-tier Publications1. Mitsyuk A. A., Lomazova I. A., van der Aalst W. M. P. Using Event Logsfor Local Correction of Process Models // Automatic Controland Computer Sciences. 2017. Vol. 51. No. 7. P. 709–723. (SCOPUS andWoS-indexed Journal Paper, Q3; author contribution: 0,9) [8]A translation of the paper in Russian: Мицюк А. А., Ломазова И.
А., вандер Аалст В. М. П. Использование журналов событий для локальной корректировки моделей процессов // Моделированиеи анализ информационных систем. 2017. Т. 24. № 4. С. 459–480. (HSE’sList Journal Paper; author contribution: 0,9) [9]2. Shugurov I. S., Mitsyuk A. A. Iskra: A Tool for Process ModelRepair // Proceedings of the Institute for System Programming. 2015. Vol.27. No. 3. P. 237–254. (HSE’s List Journal Paper; author contribution: 0,6)[10]3. Mitsyuk A.
A., Shugurov I. S. On Process Model Synthesis Based onEvent Logs with Noise // Automatic Control and Computer Sciences.4https://www.hse.ru/en/science/disscoun/council_computerscience/62016. Vol. 50. No. 7. P. 460–470. (SCOPUS and WoS-indexed Journal Paper,Q4; author contribution: 0,8) [11]A translation of the paper in Russian: Мицюк А. А., Шугуров И. С. Синтез моделей процессов по журналам событий с шумом //Моделирование и анализ информационных систем. 2014.
Т. 21. № 4. С.181–198. (HSE’s List Journal Paper; author contribution: 0,8) [12]Third-tier Publications1. Mitsyuk A. A. Non-Local Correction of Process Models UsingEvent Logs, in: Proceedings of the 2017 Ivannikov ISPRAS OpenConference. Los Alamitos : IEEE Computer Society , 2018. Ch. 2. P. 6–11.(SCOPUS-indexed Conference Paper; author contribution: 1) [13]2. Mitsyuk A. A., Lomazova I.
A., Shugurov I. S., van der Aalst W. M. P.Process Model Repair by Detecting Unfitting Fragments, in:Supplementary Proceedings of the 6th International Conference on Analysisof Images, Social Networks and Texts (AIST-SUP 2017), Moscow, Russia,July 27-29, 2017. CEUR-WS.org Vol. 1975 Aachen, 2017. Ch. 32. P. 301–313. (SCOPUS-indexed Workshop Paper; author contribution: 0,7) [14]3. Shugurov I. S., Mitsyuk A. A. Generation of a Set of Event Logswith Noise, in: Proceedings of the 8th Spring/Summer Young Researchers’Colloquium on Software Engineering (SYRCoSE 2014).
Характеристики
Тип файла PDF
PDF-формат наиболее широко используется для просмотра любого типа файлов на любом устройстве. В него можно сохранить документ, таблицы, презентацию, текст, чертежи, вычисления, графики и всё остальное, что можно показать на экране любого устройства. Именно его лучше всего использовать для печати.
Например, если Вам нужно распечатать чертёж из автокада, Вы сохраните чертёж на флешку, но будет ли автокад в пункте печати? А если будет, то нужная версия с нужными библиотеками? Именно для этого и нужен формат PDF - в нём точно будет показано верно вне зависимости от того, в какой программе создали PDF-файл и есть ли нужная программа для его просмотра.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
