Ontology-based Data Access for Extracting Event Logs from Legacy Data The onprom Tool and Methodology D. Calvanese 1 , T. E. Kalayci 1 , M. Montali 1 , S. Tinella 2 1 KRDB Research Centre for Knowledge and Data, Free University of Bozen-Bolzano (Italy) 1 2 EBITmax srl (Italy)
Table of contents 1. Introduction 2. Case Study and Motivation 3. The onprom Tool and Methodology 4. Conclusions and Future Work 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 2
Introduction
Introduction Organizations increasingly recognizing the importance of analyzing their business processes for • quality assurance • optimization • continuous improvement Process Mining [1] • The most promising and effective framework to tackle this need • It is at the intersection of model-driven engineering and data science 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 4
Introduction Insights are automatically extracted from event data to represent the footprint of process executions inside the company to [2] • discover and enrich process models • provide operational support • check compliance • analyse bottlenecks • compare process variants • suggest improvements. Plethora of process mining techniques and technologies in several application domains 1 1 http://tinyurl.com/ovedwx4 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 5
Applicability Applicability of process mining depends on two crucial factors • the availability of high-quality event data • the representation in a format that is understandable by process mining algorithms 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 6
First Setting Adopting a business process or enterprise management system that logs cases, events and corresponding attributes explicitly • XESame [3], ProMimport [4], Disco 2 , Celonis 3 , and Minit 4 support the conversion from CSV or spreadsheet files into XES • Techniques for the extraction of event logs from redo-logs of relational databases [5] • Techniques that leverages the relational technology to access the event log directly, instead of materializing it into XML [6]. 2 https://fluxicon.com/disco/ 3 http://www.celonis.de/en/ 4 http://www.minitlabs.com 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 7
Second Setting Adopting a more general management system, configuring it for its own specific needs, and combining it with domain-specific databases and other legacy data sources • Cases and events may not be explicitly stored, but instead implicitly present inside the company information system • Not a single notion of case and related events and they change depending on the perspective of interest, focus of the company, etc. • Not enough techniques, methodologies and tools that support domain experts and process analysts in a such setting • In that case • Logs are extracted manually (like extract-transform-load (ETL)) • This is a redundant, labor intensive and error prone process 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 8
Our Proposal • An approach based on conceptual modeling to semi-automatize the extraction of event logs from legacy information systems by leveraging the technique first presented in [7] • In our approach (onprom), humans only focus on the conceptual issues involved in the extraction: • Which are relevant concepts and relations? • How do such concepts/relations map to the underlying information system? • Which concepts/relations relate to the notion of case, event, and event attributes? • Once this information is provided the log extraction process is handled in a fully automatized way, leveraging the ontology-based data access (OBDA) [8, 9, 10] paradigm. 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 9
Case Study and Motivation
Case Study EBITmax 5 provides consultancy services in program management and BPM for a number of small and large enterprises. • Recently, they incorporated process mining to complement their standard consultancy services, enriching and comparing models with fine-grained insights automatically extracted from data, and accounting for how business processes are executed in reality. • They run a pilot project on the service provisioning and financial processes of Markas 6 : The analysis of Accounts Payable Process ( App ), which is used by Markas to handle payments to external suppliers and corresponding invoices. • For the internal management of the App , Markas does not employ a workflow management system, but relies on shared guidelines on how to handle payments, and on an ERP system to track the executed operations. • Markas management would like to understand if the App is executed as expected and, if not, where do deviations appear for the orders created in 2015. 5 http://www.ebitmax.it 6 http://www.markas.com/en/home.html 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 11
Traditional Methodology Other perspective? Create Y N data model Design Choose Extract Design Design Export to Do process views with per- relevant composite log view XES/CSV mining relevant spective tables views attributes Preparing Conceptual Data Model First step is preparation of the conceptual data model that accounts for the data maintained in the ERP at a higher level of abstraction • To discuss with managers and domain experts about the semantics of such data • Provides the basis to understand where and how they are stored within the ERP 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 12
Traditional Methodology Other perspective? Create Y N data model Design Choose Extract Design Design Export to Do process views with per- relevant composite log view XES/CSV mining relevant spective tables views attributes Choosing a perspective The second step consists in combining the research question with the data model, so as to choose a perspective for the analysis, and in particular deciding: 1. the subject of the analysis, i.e., which notion of case to adopt 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 13
Traditional Methodology Other perspective? Create Y N data model Design Choose Extract Design Design Export to Do process views with per- relevant composite log view XES/CSV mining relevant spective tables views attributes Choosing a perspective The second step consists in combining the research question with the data model, so as to choose a perspective for the analysis, and in particular deciding: 1. the subject of the analysis, i.e., which notion of case to adopt 2. which relevant events should be considered in the evolution of cases 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 14
Traditional Methodology Other perspective? Create Y N data model Design Choose Extract Design Design Export to Do process views with per- relevant composite log view XES/CSV mining relevant spective tables views attributes Choosing a perspective The second step consists in combining the research question with the data model, so as to choose a perspective for the analysis, and in particular deciding: 1. the subject of the analysis, i.e., which notion of case to adopt 2. which relevant events should be considered in the evolution of cases 3. which event attributes should be included 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 15
Data Model Excerpt of App PO Event RegisterInvoice Event RegisterInvoice poNo : int Timestamp: invCrTime Timestamp: invCrTime Case Case poCrTime : ts Case: is for Case: is for Invoice TD ActivePO invNo : int 0..1 0..1 0..1 0..1 tdNo : int invCrTime : ts is for subTime : ts refers to regTime : ts payTime : ts [0..1] Event PaySupplier Event PaySupplier Event SubmitOrder Event SubmitOrder Event GetTD Event GetTD Timestamp: payTime Timestamp: payTime Timestamp: subTime Timestamp: subTime Timestamp: regTime Timestamp: regTime Case: is for Case: is for Case: this Case: this Case: refers to − Case: refers to − 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 16
Traditional Methodology Other perspective? Create Y N data model Design Choose Extract Design Design Export to Do process views with per- relevant composite log view XES/CSV mining relevant spective tables views attributes Manual Construction of Data EBITmax started a fine-grained analysis of the ERP system and its underlying database to extract the desired information manually: 20th Int. Conf. on Business Information Systems 30 June 2017 D. Calvanese, T. E. Kalayci , M. Montali, S. Tinella 17
Recommend
More recommend
