Characterization of High and Stable Yield Genotype for Maize based on Multivariate stability Analysis and Prediction of Crop Yield using Ensemble Learning Techniques

Abstract

The production of maize globally surpassed the wheat and rice, because it is a staple crop in many regions of the world. Also, in addition to being directly consumed by people, maize is also used to make corn ethanol, animal food and several other products of maize, therefore the stability and high yield of maize is an extremely crucial part to promote long-term growth and food protection. This study looked at the effects of G ? E interaction on yield stability in 23 hybrids of maize in 53 distinct Indian environments. On the basis of analyses of variance, stability tests for multivariate stability parameters were carried out. In terms of all characteristics, the genotype and environment (G ? E interaction) differences were highly significant (p < 0.01), according to the pooled analysis of variance. A GGE biplot was created with the two principal components, which accounted for 66.02% and 8.51% variation in GEI for the corresponding yield per hectare. The GGE biplot and AMMI model exposed genotypes SYN916801, Bio 9682, DKC 9215, NMH 4313 as good with an indication of high mean yield and stability within the environment that were tested. By the results, we suggest breeding might boost output yield, and also the identified genotypes may be suggested for commercial farming. Also, for the creation of successful agricultural and food policy, reliable crop production estimates are essential. So, examinations were made on the machine learning techniques such as random forest and Gradient Boosting algorithm to predict crop yield responses in maize. While assessing statistical performance, GBM was found highly capable of predicting yield as compared to RF. Result showed that GBM is a good and adaptable machine learning technique for yield predictions.. KEYWORDS :Additive main effect and Multiplicative interaction, G ? E interaction, Machine learning techniques, Gradient boosting machine, Random forest.