A NOVEL METHOD WITH MULTILAYER FEED-FORWARD NEURAL NETWORK FOR MODELING OUTPUT YIELD IN AGRICULTURE

Abstract

The aim of this study was to examine energy use pattern and predict the output yield for greenhouse tomato production in Esfahan province of Iran. The data used in this study were collected from growers by using a face to face survey. The results revealed that diesel fuel (40%), chemical fertilizer (30%), electricity (12%) and human power (10%) consumed the bulk of energy. In this study, several direct and indirect factors have been identified to create an artificial neural networks (ANN) model to predict greenhouse tomato production. The final model can predict output yield based on human power, machinery, diesel fuel, chemical fertilizer, water for irrigation, seed and chemical poisons. The results of ANNs analyze showed that the (7?10?10?1)?MLP, namely, a network having ten neurons in the first and second hidden layer was the best?suited model estimating the greenhouse tomato production. For this topology, MSE of training, MSE of cross validation, RMSE, MAPE and R2 were 0.027, 0.019, 0.009, 0.98 and 96%, respec?vely. The sensi?vity analysis of input parameters on output showed that diesel fuel and seeds had the highest and lowest sensi?vity on output energy with 27% and 6%, respec?vely. Comparison between the ANN model and a Mul?ple Linear Regression (MLR) model showed that the ANN model can predict output yield relatively better than the MLR multiple model on the selected training and validation set.