Process mining techniques and tools for extracting information from event logs

U diplomskom radu dan je opis discipline dubinske analize procesa (engl. Process mining) i njegove pozicije u odnosu na znanost o podacima (engl. Data science) i znanost o procesima (engl. Process science). Fokus se stavlja na granu otkrivanja procesa (engl. Process discovery) kojoj je cilj rekonstrukcija modela procesa na temelju njegovog dnevnika događaja. Među značanijim algoritmima za tu svrhu je α-algoritam. Formalno se definiraju dnevnici događaja i Petrijeve mreže koji predstavljaju redom ulaz i izlaz α-algoritma. Obrađuju se alati Camunda i ProM uz pregled glavnih funkcionalnosti i primjera korištenja svakog. Isti se koriste u svrhu demostracije praktičnog primjera modeliranja procesa, njegove simulacije, ekstrakcije dnevnika događaja te, naposljetku, analize i usporedbe dobivenog rekonstruiranog procesa s polaznim. Detaljno je opisana korištena metodologija zajedno s vlastitom programskom implementacijom kao podrške u simulaciji procesa i ekstrakciji dnevnika događaja. Na kraju je dan praktičan pregled limitacija α-algoritma opisanom metodologijom.In the graduate thesis, an overview of process mining and its position in relation to data and process science is given. One type of process mining is of particular interest – process discovery. Its goal is to produce a process model using only the event log associated with the process. One of the most important algorithms that achieve that task - the α-algorithm, is described in great detail. A formal definition of event logs and Petri nets is given which make the input (event logs) and output (Petri nets) of the algorithm. The tools Camunda and ProM are explored. Each tool is examined alongside a practical demonstration. Both tools wih the help of a custom program implementation are used in order to provide a practical demonstration of a workflow which is made of process modeling, process simulation, event log extraction, event log analysis and process model reconstruction. The mentioned workflow is explained in detail and is later used to provide a practical overview of the limitations of the α-algorithm