Bibliographic citations
Aldave, C., (2014). Evaluación de la Precisión del Descubrimiento de Modelos de Procesos Aplicando Técnicas de Minería de Procesos [Tesis, Universidad Nacional de Ingeniería]. http://hdl.handle.net/20.500.14076/4432
Aldave, C., Evaluación de la Precisión del Descubrimiento de Modelos de Procesos Aplicando Técnicas de Minería de Procesos [Tesis]. : Universidad Nacional de Ingeniería; 2014. http://hdl.handle.net/20.500.14076/4432
@misc{renati/707226,
title = "Evaluación de la Precisión del Descubrimiento de Modelos de Procesos Aplicando Técnicas de Minería de Procesos",
author = "Aldave Castillo, Cecilia Valentina",
publisher = "Universidad Nacional de Ingeniería",
year = "2014"
}
In the next years the number of systems will increase, and it will become in millions of events that reflect the true flux of process in the Enterprise. Because of this, the current trend in retro-engineering is related to the application of the process mining which permit us obtain information about the process from the events logs. The process mining can be applied to different domains as the bank, education, health, government between others, also in one organization or between organizations according to the distribution of the process. In the 90's different groups are working to create different techniques to be implemented in the computer tools. Techniques that will allow us to obtain the models of the business process from the observed events inside of the events logs. But the first formal works appear in 1995. In 2009 starts the work force of the IEEE composed by: HP, IBM, Fujitsu, the international cabinet of consulting Gartner y Deloitte, as well as more than 20 universities among them the University of Technology of Eindhoven in Netherlands (represented by the head of technology department and the editor associated to the IEEE the professor and Doctor Wil Van Der Aalst who has published different articles about the subject), with the finality of promote the use and create the documentation about the process mining. Nowadays there are more and more tools suppliers that they are adding functionalities of process mining in their products. In this way, the scientists of the University of Technology of Eindhoven have created and continually updated the free tool of Process Mining ProM that now integrates others similar tools as EMIT, Thumb, Mison, between others, in order to make it more complete. They have created a web page to make this tool popular and there is a call to all the scientists and people who wish to try it and contribute with feedback for the continuous improvement of ProM. In relation to this, although ProM provides a support to the process mining, we still ask us: - lf it is possible to assure that the models of process generated by this tool will be really precise and clear. In other words, if in the moment that we applied the process mining from the events logs we will obtain models that represent the exact reality of these events logs or if these models will be incompletes or difficult of understand. - lf indeed, it would be advisable to employ these tools regularly, in order to support for the discovering and in consequence the improvement of process in the Enterprise and if not: ¿ What else is needed to implement ProM? ¿ What is it necessary for this tool to function well?. With the objective to answer these questions, this research tries to propose an evaluation framework. This evaluation framework will be used to evaluate the process mining tool ProM with the purpose of having an opinion and a recommendation about this representative tool. Because of this, we will begin developing a state of art which will speak about the the theories and concepts to use about the application of the process mining, the process diagrams, the event logs and the workflow patterns. We will detail an equivalence between the way of reading an event log and the way this reading will be represented through the workflow patterns by employing the algorithm Alpha by ProM. The comparison between the manual model generated by the observation of the source event log and the model that will be obtained as a result of the application of the process mining, will allow us to know if there are or not a good generation of the model and in case of failure, get to know the reasons why the generation the model was not so good. The reaserch of the generation of the models through process mining from an event log will be very detailed. The instances of models and the event log to be evaluated will have a variety of events. In other words, the number of events will be in different amounts for each process instance and these process instances will have a diversity of workflow patterns such as the following: The directly succession, the parallelism, the election, the causality between others. So the results will be more reliable to the reality. The analysis will be done by the comparison of each workflow pattern of the manual original model, which represents the event log will be used in the process mining, and the patterns inside the model generated after the application of the process mining. Also, during the study of this representative tool, we can know what are the functionalities of the current process mining tools which are best developed and which are not. The hypothesis will focus on whether the use of the process mining tool ProM has generated what really should generate or not, for this in addition to the observation method we will count the total of elements generated and employ the T student method to evaluate this hypothesis. This valídatíon will allow us contrast the result of the test with what is observed and then come to the conclusions and recommendations of the research. Finally, I would like to mention that this research has been inspired in the courses that I took in my studies of the master degree “Informatics Methods Applied to the Management of Enterprises“ in the University Pantheón Sorbonne and Paris Descartes (Paris, France 2010-2012). Also, in my work of investigation for my master memory [2] where: 1 perform the evaluation of the accuracy of a tool in the generation of the class models by applying the retro engineering of programs which have been previously generated by direct engineering. This work was made with my professor and master memory advisor: Camille Salinesi (Professor of the University Paris 1 Panthéon Sorbonne in Paris, France, Director of the research Center in informatics and Director of the Master 2 “Systems of information and of Decision“), who also provided me some information about the subject of this thesis. Similarly, the Dr. W.M.P. Van Der Aalst who helped me with the access to a real event log of a national hospital in Netherlands. As well, this thesis was also inspired by the work of retro modeling that I did in the area of Management information Systems at the bank BNP Paribas in Paris
This item is licensed under a Creative Commons License