Artificial Neural Network Use For Sweet Corn Water Consumption Prediction Depending On Cultivation Technology Peculiarities

Abstract

The goal of our study was to determine reliability of the artificial neural network methodfor prediction of the sweet corn water consumption in the irrigated conditions of the South of Ukraine. The field trials were carried out in 2014-2016 in four replications at the drip-irrigated experimental plots of the agricultural cooperative farm “Radianska Zemlia” in Kherson region. The actual water consumption of sweet corn was determined by the field measurements, and was calculated as the sum of consumed soil water, effective rainfall and irrigation water applied to the field. Artificial neural network with the architecture 3-6-1-3-1 was designed on the basis of the gathered field data within JustNN software application. After learning and training of the developed neural network we tested its reliability by the comparison of true values with predicted ones. The analysis proved high accuracy of the prediction, which is approved by the high value of the coefficient of determination (R2 ) - 0.937. We conjecture that artificial neural network algorithms are reliable enough to provide forecasts of the crops water consumption depending on the cultivation technology treatments.