Prediction of Wedelia trilobata Growth under Flooding and Nitrogen Enrichment Conditions by Using Artificial Neural Network Model

Abstract

The objective of this study is to produce multi-criteria model for the dry weight prediction of Wedelia trilobata under flooding and nitrogen conditions. Plants of W. trilobata were grown in a greenhouse, and treatments were given for two months . Growth parameters of 60 plants were used to build a numerical model. The neural network model was built using Quasi-Newton approaches that containing Broyden-fletcher-goldfarb-shanno gradient (BFGS) learning algorithm, multilayer perceptron (MLP) training algorithm and sigmoid axon transfer function along with 10 neurons at the input network, 9 neurons in the hidden layer, and 1 neuron in the output layer (10-9-1). The selection and validation of the best predictor model were based on lower values of errors and higher value of R 2 . The selected model had a higher values of R 2 = 0.90 and lower values of errors i.e (relative approximate error, RAE = 0.004, root mean square error, RMS = 0.027, mean absolute error, MAE = 0.004, mean absolute percentage error, MAPE = 0.013). Moreover, the highest rank 1 was obtained for leaf area during sensitivity analysis followed by water potential and photosynthesis ranked 2 rd and 3 th , respectively. The constructed model of W. trilobata under flooding and nitrogen conditions is the new feature in the management of invasive plant species and gives direction to control its spread.