Selection indices and support vector machines in the selection of sugarcane families

Abstract

The present study aimed to compare selection indices, namely: Smith and Hazel multiplicative, Mulamba and Mock's, and the support vector machines algorithm for sugarcane families selection. We used two datasets, from Moreira et al. (2021) and from Ferreira et al. (2022), both related to the sugarcane breeding program conducted at the Center for Sugar cane Research and Breeding at the Federal University of Viçosa, Oratórios, Minas Gerais. Both experiments were conducted in a randomized complete block design. We constructed the selection indices via mixed models approach. We adopted a selection percentage of 18% of the top families for the selection process. In both studies, we considered as explanatory traits: the number of stalks, stalks diameter and stalk height, and as the response trait the tons of stalks per hectare per family. In the dataset from Ferreira et al. (2022), the support vector machine was a better approach to select sugarcane families by learning from the data after multivariate simulation. Whereas in the dataset from Moreira et al. (2021), using similar methodology, lower performance for support vector machines was obtained. Keywords: Synthetic data. Indirect selection. Yield prediction. Machine learning. BLUP