skip to main contentskip to main menuskip to footer Universität Bielefeld Play

Kolloquium des ZeSt

 

Dienstag, 25.10.2022, 12-13 Uhr

David Winkelmann
Universität Bielefeld

Integrated and Robust Storage Assignment: An E-Grocery Retailing Business Case

In this talk, we deal with a storage assignment problem arising in a fulfilment centre of a major European e-grocery retailer. The centre can be characterised as a hybrid warehouse consisting of a highly efficient and partially automated fast-picking area designed as a pick-and-pass system with multiple stations and a picker-to-parts area. The storage assignment problem considered in this paper comprises the decisions to select the products to be allocated to the fast-picking area, the assignment of the products to picking stations and the determination of a shelf within the assigned station. The objective is to achieve a high level of picking efficiency while respecting station workload balancing and precedence order constraints. We propose to solve this three-level problem using an integrated MILP model. In computational experiments with real-world data, we show that using the proposed integrated approach yields significantly better results than a sequential approach in which the selection of products to be included in the fast-picking area is solved before assigning station and shelf. Furthermore, we provide an extension to the integrated storage assignment model that explicitly accounts for within-week demand variation. In a set of experiments with day-of-week-dependent demands we show that while a storage assignment that is based on average demand figures tends to exhibit a highly imbalanced workload on certain days of the week, the augmented model yields robust storage assignments that are well balanced on each day of the week without compromising the quality of the solutions in terms of picking efficiency.



Dienstag, 08.11.2022, 12-13 Uhr

Hendrik van der Wurp
TU Dortmund

Introducing Regularisation to Generalised Joint Regression Modelling for Count Data

When modelling the bivariate outcome of football matches and other sports, many different approaches regarding dependency have been investigated. We propose the use of copula regression via the powerful GJRM (Generalised Joint Regression Models) framework in R by Giampiero Marra and Rosalba Radice and present its use for modelling match results. Motivated by the application to football and FIFA World Cups in particular, we introduce two types of useful penalties. The first tackles a very specific issue occurring in sport tournaments and leagues (or other competitive situations), while the second is a Lasso-approximation yielding general sparsity.



Dienstag, 22.11.2022, 12-13 Uhr

Dr. Jennifer Pohle
Universität Potsdam

Markov-switching conditional logistic regression for animal movement data


Animals’ habitat and resource use is of central interest in ecology. A popular statistical framework to study fine-scale movement and habitat selection based on telemetry data and conditional logistic regression is the so-called integrated step selection analysis. It has also been successfully applied to detect interactions (such as avoidance or attraction) between simultaneously tracked individuals. Animals’ preferences, habitat selection, and movement patterns might, however, depend on their usually unobserved behavioural modes (such as resting and foraging). Ignoring such behavioural states in the analyses might lead to biased results and possible erroneous conclusions. In this talk, a Markov-switching conditional logistic regression model is introduced which allows for underlying latent state dynamics and corresponding state-dependent movement and selection coefficients. The proposed method is illustrated using simulations and a real data case study on simultaneously tracked bank voles. Furthermore, its connection to spatial point processes is discussed.



Dienstag, 06.12.2022, 12-13 Uhr

Dr. Marc Hüsch
Centrum für Hochschulentwicklung

CHE Ranking & CHE Hochschuldaten – mit Datenerhebungen, statistischen Analysen und interaktiven Visualisierungen die Vielfalt der Hochschullandschaft begreifbar machen

Wie eine aktuelle Auswertung des Centrums für Hochschulentwicklung (CHE) zeigt, wächst die Vielfalt der Studienangebote in Deutschland immer weiter: im HRK Hochschulkompass wurden im Jahr 2021 über 21.000 unterschiedliche Studienangebote verzeichnet (https://www.che.de/download/studiengaenge-2021/). Studieninteressierte können somit aus einer Vielzahl an unterschiedlichen Angeboten wählen. Um bei dieser Vielfalt für eine Orientierung zu sorgen, stellt das CHE mit dem CHE Hochschulranking eine umfangreiche Orientierungs- und Informationsplattform für die Studienwahl zur Verfügung. Dazu erhebt das CHE unter anderem vielfältige Daten von den Fachbereichen und führt Studierendenbefragungen durch, in der die aktuellen Studierenden Auskunft über ihre Erfahrungen zu unterschiedlichen Themen ihres Studiums geben können. Nach den Erhebungen werden die Daten vom CHE statistisch ausgewertet und im ZEIT Studienführer sowie im Online-Ranking unter https://ranking.zeit.de/che/de/ veröffentlicht. Im Vortrag werden die methodische Vorgehensweise sowie spezielle Herausforderungen bei den Erhebungen und Auswertungen für das CHE Ranking vorgestellt, wobei ein besonderes Augenmerk auf der Studierendenbefragung liegt. Neben den Informationsangeboten für Studieninteressierte stellt das CHE mit dem neuen Portal CHE Hochschuldaten unter www.hochschuldaten.de zudem auch ein Angebot für Medien, politische Akteure und andere Interessierte zur Verfügung, die sich für den Status Quo und für die Entwicklung der Hochschullandschaft in Deutschland und in den einzelnen Bundesländern interessieren. Dabei werden Daten des Statistischen Bundesamtes sowie aus CHE-eigenen Erhebungen mit Hilfe von interaktiven Visualisierungen veranschaulicht. So kann unter anderem die Entwicklung der Studierendenzahlen in einzelnen Studienfächern sowie Analysen zur Wohn- und Lebenssituation von Studierenden eingesehen werden. Zum Einsatz kommen dabei Daten-Dashboards auf Basis von R Shiny sowie interaktive Grafiken auf Basis des Online-Tools Datawrapper. Im Vortrag werden die Möglichkeiten des Portals und die verwendeten Tools vorgestellt.



Dienstag, 20.12.2022, 12-13 Uhr
- Termin fällt aus -



Dienstag, 17.01.2023, 12-13 Uhr

J.-Prof. Dr. Julian Hinz
Universität Bielefeld

Brothers in arms: The value of coalitions in sanctions regimes

This paper examines the impact of coalitions on the economic costs of the 2012 Iran and 2014 Russia sanctions. By estimating and simulating a quantitative general equilibrium trade model under different coalition set-ups, we (i) dissect welfare losses for sanction-senders and target; (ii) compare prospective coalition partners and; (iii) provide bounds for the sanctions potential — the maximum welfare change attainable — when sanctions are scaled vertically, i.e. across sectors up to an embargo, or horizontally, i.e. across countries up to a global regime. To gauge the significance of simulation outcomes, we implement a Bayesian bootstrap procedure that generates confidence bands. We find that the implemented measures against Iran and Russia inflicted considerable economic harm, yielding 32 – 37% of the vertical sanctions potential. Our key finding is that coalitions lower the average welfare loss incurred from sanctions relative to unilateral implementation. They also increase the welfare loss imposed on Iran and Russia. Adding China to the coalition further amplifies the welfare loss by 79% for Iran and 22% for Russia. Finally, we quantify transfers that would equalize losses across coalition members. These hypothetical transfers can be seen as a sanctions-equivalent of NATO spending goals and provide a measure of the relative burden borne by coalition countries.



Dienstag, 31.01.2023, 12-13 Uhr

Ferdinand Stoye
Universität Bielefeld

Extensions of hidden Markov models for high resolution ecological data

Hidden Markov models have become a standard tool in modeling animal movement data over the past several years. A central assumption of HMMs is the conditional independence of observations over time, given the states of the model. When employing modern methods in the generation of (animal movement) data, the acquisition of several observations per second is now possible. However, such high-resolution data lead to a violation of conditional independence. For example, while a bird is in a turn, its turning angle is not independent of the previous angle (given the states). Thus, it is very unlikely to change its direction much within a short time. In this talk, we address the modeling of autocorrelation in the state-dependent process of HMMs. We investigate the properties of such models in a simulation study and then apply them to high-resolution movement data of a sea tern. Of particular interest in this context is the specification of appropriate distributions as well as the choice of the degree in the autoregression.
 


back to top