INTEGRATED GENETIC COMPONENTS AND MACHINE LEARNING APPROACHES FOR BETTER SELECTION OF TRAITS IN BREEDING OF MELON UNDER HIGH TUNNEL CULTIVATION CONDITION

Morphologic depiction and genetic components analysis of fruit traits and yield were done in 10 Iranian melon landraces. Significant differences were detected by analysis of variance for all measured traits, but year effect was not significant, indicating that the significant difference was not observed year to year and was not effective on the phenotypic values. Phenotypic variances were higher than genotypic variances for most of the traits. Number of fruit and flesh diameter (74% and 70%) had the highest heritability between traits respectively. Correlation coefficients indicate that the phenotypic correlation coefficients were higher than the genotypic correlation coefficients in studied traits. Yield possessed positive and highly significant correlation with fruit weight, flesh diameter, cavity diameter. Cluster analysis by ward method grouped genotypes into the two groups. Random forest analysis was done on 60 samples of genotypes, Variable importance by random forest approach revealed flesh diameter and cavity diameter depicted as effective variables in identifying yield variation. SVM graph showed cavity diameter and fruit length seperate out genotype sefidak with high confidence from the others.