Regression methods estimation of genotype yield value in plant breeding
2019
Koyuncu, Bengu | Gok, Murat
The aim of scientific research is to reach general results from the observations and experiments of the studies. Together with the developing technologies, these results are recorded digitally, and these records form big data stacks. The process of processing these masses into meaningful information began in the 1950s and the concept of data mining emerged. Data mining, which is used in forecasting or decision making processes, now finds its place in forecasting agricultural activities. The basis of the plant breeding studies is based on the comparison of the desired phenotype and genotype properties according to the efficiency and environmental conditions. Various statistical package programs are used in the evaluation of these results. These programs do not fully meet the analysis and reporting capabilities required for a breeder’s genotype selection. In this study, the yield of the genotype was estimated from 12 locations with a total of 1153 yields of 24 replicates of 24 genotypes. In addition to the linear regression used in plant breeding, sequential Minimal Optimization (SMO), Nearest k–Neighbor (k–EYK), Random Forest (RO) methods were selected from the methods of machine learning. The success of selected methods was compared according to the mean square root of the mean square error and the mean absolute error metrics. RO has shown higher performance than the other three methods and it has been proposed to use with the linear regression method used in plant breeding programs.
显示更多 [+] 显示较少 [-]