Modelling of jute production using artificial neural networks

Neural network models, trained by back propagation, were developed to predict the development of jute using previously obtained experimental field data. The models were based on a common structure and developed using data from four sets of experiments conducted in 2006 and 2007 using two different varieties of jute. The models consisted of four-layered networks and a large number of neurons. The six input variables were represented by six neurons; Julian day, solar radiation, maximum temperature, minimum temperature, rainfall, and type of biomass. The output variable, represented by a single neuron, was plant dry matter. The models had two hidden layers with 9 and 5 neurons. The two sets of experiments conducted in 2006 were used for training the models and with two sets of experiments conducted in 2007 used for validation. The models accurately predicted jute production. They could be used to predict production at different locations and could be used to predict yield of other field crops if trained properly. (C) 2009 IAgrE. Published by Elsevier Ltd. All rights reserved.