JHJulia Heger

Julia Heger

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Julia Heger has a total of 4 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Group-constrained assortment optimization under the multinomial logit model

    by Julia Heger, Robert Klein

    We study an assortment problem under the multinomial logit model with two new types of group constraints that are motivated by a joint project with the German car manufacturer BMW. Under group constraints, products are either attributed to exactly one group or to several groups at once and there is either a bound on the number of products offered per group or on the number of groups from which products are offered. We formulate both optimization problems as binary fractional linear program and provide reformulations that can be solved using state-of-the-art solvers. Finally, we conduct a numerical study and find that all instances of the products-per-group constrained problem as well as small to medium size instances of the number-of-offered-groups constrained problem can be solved within fractions of a second, whereas large instances of the latter problem might take some seconds to be solved.


  • Open Access

    Assortment optimization: a systematic literature review

    by Julia Heger, Robert Klein

    Assortment optimization is a core topic of demand management that finds application in a broad set of different areas including retail, airline, hotel, and transportation industries as well as in the healthcare sector. Hence, the interest in research on assortment optimization has grown rapidly in recent years. However, the sheer number of publications on the topic of assortment optimization makes it difficult to keep track of all available approaches proposed in the literature. In this paper, we systematically review state-of-the-art studies on assortment optimization. We assemble an extensive literature overview by strategically searching for pre-defined keywords within leading scientific databases. The resulting literature is grouped by a proposed taxonomy that captures properties related to the optimization problem itself, the modelled customer behaviour, and the solution concept applied for solving the problem at hand. For each group, we provide an overview of the corresponding literature and analyse it based on a proposed selection of key factors.


  • Open Access

    Detection of interacting variables for generalized linear models via neural networks

    by Yevhen Havrylenko, Julia Heger

    The quality of generalized linear models (GLMs), frequently used by insurance companies, depends on the choice of interacting variables. The search for interactions is time-consuming, especially for data sets with a large number of variables, depends much on expert judgement of actuaries, and often relies on visual performance indicators. Therefore, we present an approach to automating the process of finding interactions that should be added to GLMs to improve their predictive power. Our approach relies on neural networks and a model-specific interaction detection method, which is computationally faster than the traditionally used methods like Friedman’s H-Statistic or SHAP values. In numerical studies, we provide the results of our approach on artificially generated data as well as open-source data.


  • Open Access

    Detection of Interacting Variables for Generalized Linear Models via Neural Networks

    by Yevhen Havrylenko, Julia Heger

    The quality of generalized linear models (GLMs), frequently used by insurance companies, depends on the choice of interacting variables. The search for interactions is time-consuming, especially for data sets with a large number of variables, depends much on expert judgement of actuaries, and often relies on visual performance indicators. Therefore, we present an approach to automating the process of finding interactions that should be added to GLMs to improve their predictive power. Our approach relies on neural networks and a model-specific interaction detection method, which is computationally faster than the traditionally used methods like Friedman H-Statistic or SHAP values. In numerical studies, we provide the results of our approach on artificially generated data as well as open-source data.