Detection of interacting variables for generalized linear models via neural networks
Research output: Contribution to journal › Journal article › Research › peer-review
Standard
Detection of interacting variables for generalized linear models via neural networks. / Havrylenko, Yevhen; Heger, Julia.
In: European Actuarial Journal, 2024.Research output: Contribution to journal › Journal article › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - JOUR
T1 - Detection of interacting variables for generalized linear models via neural networks
AU - Havrylenko, Yevhen
AU - Heger, Julia
N1 - Publisher Copyright: © 2023, The Author(s).
PY - 2024
Y1 - 2024
N2 - 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.
AB - 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.
KW - Generalized linear model
KW - Insurance claims prediction
KW - Interaction detection
KW - Model interpretability
KW - Neural network
UR - http://www.scopus.com/inward/record.url?scp=85175378985&partnerID=8YFLogxK
U2 - 10.1007/s13385-023-00362-4
DO - 10.1007/s13385-023-00362-4
M3 - Journal article
AN - SCOPUS:85175378985
JO - European Actuarial Journal
JF - European Actuarial Journal
SN - 2190-9733
ER -
ID: 372718973