An artificial intelligence approach for modeling volume and fresh weight of callus – A case study of cumin (Cuminum cyminum L.)

Abstract

Cumin (Cuminum cyminum Linn.) is valued for its aroma and its medicinal and therapeutic properties. A supervised feedforward artificial neural network (ANN) trained with back propagation algorithms, was applied to predict fresh weight and volume of Cuminum cyminum L. calli. Pearson correlation coefficient was used to evaluate input/output dependency of the eleven input parameters. Area, feret diameter, minor axis length, perimeter and weighted density parameters were chosen as input variables. Different training algorithms, transfer functions, number of hidden nodes and training iteration were studied to find out the optimum ANN structure. The network with conjugate gradient fletcher-reeves (CGF) algorithm, tangent sigmoid transfer function, 17 hidden nodes and 2000 training epochs was selected as the final ANN model. The final model was able to predict the fresh weight and volume of calli more precisely relative to multiple linear models. The results were confirmed by R2≥0.89, R(i)≥0.94 and T value ≥0.86. The results for both volume and fresh weight values showed that almost 90% of data had an acceptable absolute error of ±5%.