Simulation of Yield and Water Productivity of Cucumber Plant Using Artificial Neural Network

 In order to simulate the yield and water productivity of cucumber plant (Cucumis sativus L.), an experiment was conducted in the form of a completely randomized block design with three irrigation levels of 100, 85 and 75% of the water requirement in two growing seasons during 2017 and 2018 and using perceptron neural networks (MLP) and support vector machine (SVM) methods were used and finally, to select the appropriate and optimal model, the indices of explanatory coefficient, mean squared error and normalized mean squared error were used. The amount of irrigation water, number of leaves on the plant, temperature, evaporation rate and relative humidity were selected as input data and 60%, 20% and 20% of the total data were allocated for training, validation and testing of the model, respectively. The results showed that the MLP neural network with the inputs of irrigation water and number of leaves was more accurate in simulating fruit yield and water productivity in cucumber plants with an explanation coefficient of 0.92 and 0.86, respectively. The results of the sensitivity analysis indicated that the irrigation water input parameters are the most important effective parameters on the water consumption efficiency model and cucumber fruit yield with sensitivity coefficients of 0.9 and 0.86, respectively.