How to automate master data management with machine learning?
In recent years, data quality has become a crucial issue for many businesses. The purpose of this paper is to summarize a project that aims at finding and researching approaches that resolve typical master data quality issues of a manufacturing company.
Pascal Schaller, 2021
Bachelor Thesis, Manufacturer
Betreuende Dozierende: Kaspar Riesen
Keywords: Master data management, Data quality, Data analysis
Views: 17 - Downloads: 3
Insufficient training, undefined or unclear processes, and the decentralization of subsidiaries in the past result in challenges for the company’s master data management (MDM) and master data quality today. Due to the described sources of the MDM problem, three goals were determined. First, an analysis of the article master data has to be conducted. Second, an identification and recommendation of MDM best practices has to be made. Third, a recommendation of how to proceed, based on the results of the two other objectives, will be given.
A taxonomy of so-called dirty data was used to develop diverse data analysis methods. The methods were employed for the analysis of the article master data. Best practices were identified and suggested based on the assessment of the company’s MDM maturity. To this end, a maturity model for the achievement of the second objective is used. To achieve the third objective, the maturity model was used to suggest an implementation roadmap for future MDM. The company can now follow this roadmap in order to solve and prevent master data quality issues. Finally, the data analysis methods have been connected to the initial model to show how the methods can be used to mature in MDM.
The three objectives resulted in a set of reactive and proactive approaches that can be used to solve diverse data quality issues the company currently faces in their master data. By means of the developed data analysis methods the company can now identify data quality issues as a reactive approach. Moreover, the assessment of the company’s MDM maturity along with the identification of MDM best practices is also helpful. Actually, both can be used as a proactive approach and can be furthermore used as basis for a roadmap to unify reactive and proactive approaches. The data analysis methods developed identified 3.1% of all master data articles as being of insufficient data quality. Also potential patterns of insufficient data quality were found such as wrongful deactivations of data objects. The assessment of the MDM maturity identified various MDM capabilities that should be implemented by the company to operate MDM with best practices. Regarding the maturity model the company is missing 26 more capabilities to operate MDM with best practices. The roadmap proposes four projects based on the identified capabilities that mature the MDM of the company. An implementation of these projects might solve and prevent master data quality issues in the future.
Studiengang: Business Information Technology (Bachelor)
Fachbereich der Arbeit: Business Information System & IT-Management