Veebipõhine tööriist ennetava protsessijälgimise uurimise toetemiseks
Files
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Äriprotsesside ennustav seire üritab ennustada hetkel toimuva äriprotsessi lõpptulemust, kestvust või maksumust, kasutades selleks sündmuste logides leiduvaid eelnenud sündmuseid. Käesolevas magistritöös kirjeldame veebirakenduse funktionaalsust (Nirdizati Research), mis võimaldab kasutajatel leida konkreetse sündmuste logi jaoks parima ennustava mudeli. Seda mudelit saab seejärel kasutada käimasolevate protsessidele ennustuste tegemiseks.Nirdizati Research lubab kasutajal teha mudeli äriprotsessi järele jäänud aja, tulemuse või järgmise sündmuse ennustamiseks. Rakendus pakub mudeli tegemiseks erinevaid konfigureerimise võimalusi ning võimaldab visuaalselt võrrelda genereeritud mudeleid. Rakendust on hinnatud haigla raviprotsessi sündmuste logi põhjal. Jõudluse hindamiseks on kasutatud finantsinstitutsiooni taotlusprotsessi logifaili.
Predictive Process Monitoring aims at predicting the outcome, time or cost of an ongoing execution of a business process using past process executions recorded in event logs. In this Master's Thesis, we describe the functionality of a web application (Nirdizati Research) that can be used to find a predictive model extracted from a given event log that can be used at runtime for making predictions on ongoing cases. Nirdizati Research allows the user to create a model to predict the remaining time, the outcome and the next activity of an ongoing process execution. The application offers various configuration options in a way that different predictive models can be rated using various techniques and methods. There are also options to compare the created prediction models using different metrics. The tool has been evaluated on an authentic event log concerning the treatment process of sepsis patients in a hospital. Furthermore, the performance of the tool has been measured using a real-life log pertaining to the application process in a financial institute.
Predictive Process Monitoring aims at predicting the outcome, time or cost of an ongoing execution of a business process using past process executions recorded in event logs. In this Master's Thesis, we describe the functionality of a web application (Nirdizati Research) that can be used to find a predictive model extracted from a given event log that can be used at runtime for making predictions on ongoing cases. Nirdizati Research allows the user to create a model to predict the remaining time, the outcome and the next activity of an ongoing process execution. The application offers various configuration options in a way that different predictive models can be rated using various techniques and methods. There are also options to compare the created prediction models using different metrics. The tool has been evaluated on an authentic event log concerning the treatment process of sepsis patients in a hospital. Furthermore, the performance of the tool has been measured using a real-life log pertaining to the application process in a financial institute.