2023

Procedia Computer Science

Journal year: 2023 | Journal volume: 225

scientific article / paper

english

EN A dataset shift takes place in a situation, where the joint distribution of inputs and outputs differs between the stages of training, testing, and using predictive models. If the distribution of current data for the implemented forecasting model changed significantly compared to the distribution of data used to develop it, them it could lead to its incorrect operation. The aim of the study was to compare the properties of two indicators, the Univariate Predictive Accuracy Index (UPAI) and Multivariate Predictive Accuracy Index (MPAI). The research procedure was carried out in 2 scenarios. The first involved a comparison of UPAI and MPAI for the distributions of categorical variables, while the second for distributions of continuous variables. The obtained MPAI values for both scenarios were summarized and compared with the UPAI and PSI values calculated and published in our previous article [1]. The results of the experiment proved that basing the decision on the need to calibrate the model or build a completely new model on the basis of MPAI, a measure that takes into account the multivariate distribution of variables, is superior to one-dimensional measures

3342 - 3351

10.1016/j.procs.2023.10.328

https://www.sciencedirect.com/science/article/pii/S1877050923014862?via%3Dihub

27th International Conference on Knowledge Based and Intelligent Information and Engineering Sytems (KES 2023), 6-8.09.2023, Athens, Grecja

CC BY-NC-ND (attribution - noncommercial - no derivatives)

open journal

final published version

12.2023 (Date presumed)

at the time of publication

5

70