Unterstützung der Leistungsdiagnostik im Leistungs- und Hochschulsport durch maschinelles Lernen am Beispiel der Ausdauerdiagnostik

Research output: Book/ReportDissertationsResearch


  • Benedikt Malecki

Research units


Athletes chasing after new records – backed by the insights of sports science – outperform the boarders of humane capability. Those insights are based on the collection and analysis of diverse data.
In the years to come in sports science – similar to the advances in medical science – personalized approaches to support decision-making in the field of endurance diagnostics and endurance training are getting more and more important.
To follow such personalized approaches, processes of data organisation and data analyses with artificial intelligence as well as the communication of the- se insights have to be implemented.
At present, the use of such common approaches is scarce.
Based on the above-mentioned requirements the data warehouse process was implemented as the higher-ordered process, the subprocess for the data ana- lysis by the CRoss Industry Standard Process for Data Mining (CRISP-DM). For this purpose a server-client-system with a JSON-API and an underlying relational database was developed to avoid the high costs of aquirement and maintenance of heavy business intelligence systems. Based on a two-step ma- chine learning model a division of athletes into two endurance performance clusters was accomplished by using a hierarchical clustering algorithm. Af- terwards rules were found by using a decision tree to identify parts of the physiological structures respectively their characteristics within the clusters.
Therefore the built model gives insights in the physiological structures of endurance performance clusters. Thus the model could be used to support trainers in endurance diagnostics and in planning endurance training. Future work has to be done to test the model in the daily diagnostic and training routine.
Original languageGerman
Place of PublicationKöln
PublisherDeutsche Sporthochschule Köln
Number of pages180
Publication statusPublished - 2021

ID: 6202559

View graph of relations