Data Science and Artificial Intelligence

As the field of Data Science matures, we need to define a standard approach how to implement Data Science and AI projects in practice.

The DASC-PM model defines a standard approach for data science projects, read the full paper here DASC-PM v.1.0 - Ein Vorgehensmodell für Data-Science-Projekte (in German)

Companies in particular face many challenges implementing AI in their organization. In many cases, technical infrastructure and talent are major hurdles but even if these are available, organizational and business aspects are even more challening in practical applications. Identifying suitable use-cases is of critical importance for companies to derive value from AI. The “Enterprise AI Canvas” Journal Preprint offers a structured approach and guidance how to start an AI project in a company.

Auf deutsch: Podcast Interview “Ist jetzt eigentlich alles AI?” mit dem IT Manager Podcast

Algorithms

Data Science and AI are driven by the underlying algorithms that are used to transform data into predictions.

The NeuroBayes neural network package (developed with Prof. Dr. M. Feindt) is a sophisticated 3-layer feed-forward network that can be used either for classification or for regression. It’s main advantages are that it does not overtrain and, in case of regression, can predict entire probability density distributions. These distributions include all information about the predicted event, including the uncertainty.

Most machine learning algorithms are “black boxes” in the sense that individual predictions cannot be understood in detail. The novel algorithm Cyclic Boosting Journal Preprint (with Dr. F. Wick and Prof. Dr. M. Feindt) allows to calculate accurate precisions where the underlying “factors” for each prediction can fully explain how individual predictions were made.

AI in Education

I’m part of the AI in Education initiative at IUBH where we focus on finding novel ways how AI technologies can aid the learning experience. Recently, the contribution “On Demand Tutoring in Distance Learning” by two of my Master’s students was accepted at the conference “Educational Research (Re)connecting Communities” (ECER2020) - though unfortunately, the conference itself was cancelled due to Covid-19.

Please get in touch if you would like to start a thesis topic in this area!

Together with Prof. A. Hollstein, we also start a survey about the digitalization in German schools EduC

Selected Publications

Data Science & AI

U. Kerzel (2022) Demand Models For Supermarket Demand Forecasting, International Journal of Supply and Operations Management, DOI: 10.22034/IJSOM.2022.109044.2145
F. Wick, U. Kerzel, M. Hahn, M. Wolf, T. Singhal, D. Stemmer, J. Ernst, M. Feindt (2021) Demand Forecasting of Individual Probability Density Functions with Machine Learning, SN Oper. Res. Forum 2, 37 (2021). https://doi.org/10.1007/s43069-021-00079-8
U. Kerzel (2020) Enterprise AI Canvas: Integrating Artificial Intelligence into Business, Applied Artificial Intelligence. DOI: 10.1080/08839514.2020.1826146
U. Kerzel, S. Horstmann, M. Horn, A. Hollstein (2020) On Demand Tutoring in Distance Learning, accepted at ECER 2020 (cancelled due to corona pandemic)
F. Wick, U. Kerzel, M. Feindt (2019) Cyclic Boosting - an explainable machine learning algorithm, In 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA) (pp. 358-363). IEEE.
M. Feindt, U. Kerzel (2006) The NeuroBayes neural network package, Nucl.Instrum.Meth.A559:190-194. DOI: 10.1016/j.nima.2005.11.166

Fundamental Physics

M. Adinolfi et al. [LHCb RICH Group] (2013) Performance of the LHCb RICH detector at the LHC, Eur. Phys. J. C. 73:2431. DOI: 10.1140/epjc/s10052-013-2431-9
R. Aaij et al. [LHCb Trigger Group] (2013) The LHCb Trigger and its Performance, Journal of Instrumentation 8 P04022. arXiv: 1211.3055
S. Chatrchyan et al. [CMS Collaboration] (2012) A New Boson with a Mass of 125 GeV Observed with the CMS Experiment at the Large Hadron Collider, Science 338, 1569-1575. DOI: 10.1126/science.1230816
R. Aaij et al. [LHCb Collaboration] (2012) Measurement of the effective $B_s^0 \rightarrow K^+ K^-$ lifetime, Phys. Lett. B 716, 393. arXiv: 1207.5993
M. Adinolfi et al. (2009) Performance of the LHCb RICH photo-detectors and readout in a system test using charged particles from a 25 ns-structured beam, Nucl.Inst.Meth.A603:287-293. DOI: 10.1016/j.nima.2009.02.013

Research in Data Science and AI