Predicting rice phenotypes with meta and multi-target learning

• DOI: 10.1007/s10994-020-05881-9
• Publication Date: 2020-08-02T21:14:36Z
• Authors: Oghenejokpeme I. ...; Nickolai N. Alexa...; Ross D. King
• Keywords: Feature (linguistics)

Abstract The features in some machine learning datasets can naturally be divided into groups. This is the case with genomic data, where grouped by chromosome. In many applications it common for these groupings to ignored, as interactions may exist between belonging different However, including a group that does not influence response introduces noise when fitting model, leading suboptimal predictive accuracy. Here we present two general frameworks generation and combination of meta-features feature are present. Furthermore, make comparisons multi-target learning, given one typically interested predicting multiple phenotypes. We evaluated approaches on rice dataset regression task predict plant phenotype. Our results demonstrate there use cases both meta approaches, overall, they significantly outperform base case.