Investigation Of Genetic Diversity Of Different Rapeseed (brassica napus l.) Genotypes And Yield Prediction Using Machine Learning Models

Abstract

Background Seed yield is controlled by additive and non-additive effects of genes, so predicting seed yield is one of the most important goals of rapeseed breeding in agricultural research. However, there is less information about the yield estimation of canola using neural network. In this research, three models of Multi-Layer Perceptron (MLP) neural network, Radial Basis Function (RBF) neural network and Support Vector Machine (SVM) were used to predict rapeseed yield. Network training was performed using phenological, morphological, yield and yield components, as well as data obtained from molecular markers of 8 genotypes and 56 hybrids. Results The obtained from the comparison of the efficiency of the models showed that the MLP model was able to predict the hybrid yield with the RMSE, MAE and R2 equal to 226, 183 and 92% and the use of phenotypic data as model inputs in direct crosses with the highest accuracy. In the genetic evaluation section, according to the indicators obtained, it was found that molecular study is a powerful tool that can provide valuable information to the breeder. The results showed that among the 40 primers investigated, the ISJ10 primer had more resolving power than the other primers. Conclusions The use of molecular and phenotypic data as input data in the model showed that the MLP model had a lower error value in terms of RMSE and MAE and a higher R2 than direct crosses in predicting the performance of reciprocal crosses. The proposed neural network model makes it possible to estimate the performance of each of the hybrids of the parents studied before crossing, which helps the breeder to focus on the best possible hybrids.