Article title: Machine learning predictions of concentration-specific aggregate hazard scores of inorganic nanomaterials in embryonic zebrafish
Authors: C. Gousiadou, R. L. Marchese Robinson, M. Kotzabasaki, P. Doganis, T. A. Wilkins, X. Jia, H. Sarimveis and S. L. Harper
Journal: Nanotoxicology
Bibliometrics: Volume 15, Number 4, pages 446–476
DOI: https://doi.org/10.1080/17435390.2021.1872113
The following statement in the published version of this manuscript is incorrect and was regrettably not fixed during proof-reading:
“Whilst the RMSEcv was actually not improved, this was the exception and, for all subsequent modeling on different data subsets and endpoints, ensemble modeling appeared to improve upon the base models, as can be seen in Tables 2, 3 and 5”
An accurate summary of these results has been updated to:
“As can be seen from Tables 2, 3 and 5, the ensemble modelling approaches sometimes showed improved performance, but this was not consistent across all of the different kinds of modelled data or test sets considered and, in some cases, this was not entirely consistent in terms of the different performance statistics computed. Nonetheless, the ensemble modelling approach was found to perform better than or comparably to all of the base models, in terms of all statistics, for four out of the six (pseudo-)external test sets.”