Application of hierarchical grouping and machine learning models for classification of genotypes of Nicotiana tabacum L. with based on morphological characteristics

Abstract

The maintenance of the variability of plant genetic resources is important both for conservation purposes and for the expansion of the genetic bases of the species. Machine learning is an additional tool with great potential that has been used in the search for solving various agricultural challenges. In this article we study the genetic divergence between tobacco genotypes by multivariate methods and verify the efficiency of computational algorithms, aiming at the identification of promising genotypes. Through multivariate analyses, three divergent groups were formed. Both approaches had high accuracy in classification. An artificial neural network with a hidden layer containing three neurons obtained an accuracy of 0.98, while the Random Forest accuracy was 0.98 and the decision tree 0.96. These values demonstrate how powerful these tools are and reinforce their use in studies aimed at maintaining the germplasm bank and also in tobacco conservation and genetic improvement programs in the Recôncavo region of Bahia.