Soft computing-based approach on prediction promising pistachio seedling base on leaf characteristics

Fruit trees breeding is a time-consuming process. It can save time and cost in a fruit breeding program if promising genotypes can be predicted at the early stages of vegetative growth. In the current study, Artificial neural networks analysis (ANNs) has been used to predict promising pistachio with large nuts and green kernels based on leaves characteristics at the juvenile stage. Eight morphological traits related to leaf properties of 95 pistachio genotypes in a segregating population were used to predict the number of dry nuts per ounce (N-po) and kernel color classification using radial basis function (RBF). The results of N-po modeling using RBF showed that the root mean square errors (RMSE) for the training and testing phases are 0.28 and 0.37, respectively. R-2 of 99% prediction also indicated that N-po could be estimated from leaf characteristics. Besides, the results of kernel color classification based on leaf characteristics also showed that the RBF is able to distinguish the pistachio kernel color with 98.95 per cent accuracy. The results also showed that the terminal leaflet apex (TLA) is the most critical leaf characteristic in the detection of N-po and kernel color. Based on N-po and green kernel the PIS-41, PIS-46, PIS-57, PIS-62, and PIS-67 genotypes were promising and can be used future breeding program.