Description of the relationships between different plant characteristics in soybean using multivariate adaptive regression splines (MARS) algorithm

Abstract

The aim of this study was to reveal the relationships between several morphological characteristics of the soybean (Glycine max (L.) Merr.) plants in the year 2014. For this aim, plant height (PH), first pod height (FPH), branch number (BN), number of nodes (NN), pod number per plant (PNP), seed number per pod (SNP), 1000-seed weight (1000SW), yield per decare (YD) and harvest index (HI) were measured. Five different MARS models were developed for the plant height, first pod height, pod number, harvest index and yield per decare characteristics. The constructed models were evaluated based on the criteria of minimum generalized cross-validation (GCV), SDratio, RMSE, AIC, AICc and maximum coefficient of determination (R2) in predictive performance. The R2 values of the MARS models were determined to be 0.902, 0.924, 0.949, 0.987 and 0.998, respectively. For the prediction of PH, FPH and HI, the second degree interaction model was determined to be the most suitable model. For predicting PNP and YD, the third degree interaction MARS model was determined to be the best model. The dependent variables considered here was predicted with a high accuracy by all models established with the MARS algorithm. As a result, application of the MARS algorithm may allow plant breeders to obtain influential clues in selecting promising soybean varieties. © 2020, Pakistan Agricultural Scientists Forum. All rights reserved.