Application of multi-layer adaptive neuro-fuzzy inference system for estimation of greenhouse strawberry yield

In this study adaptive neuro-fuzzy inference system (ANFIS) was employed to forecast greenhouse strawberry yield on the basis of different combination of energy inputs. Also, an artificial neural networks (ANNs) model was developed and generalized to compare their results with proposed ANFIS model. Data for the present study was randomly collected from 33 greenhouses from Guilan province, Iran. Energy inputs used in strawberry cultivation included labor, chemical fertilizers, diesel fuel, machinery, biocides, electricity, natural gas and water for irrigation which were all selected as input parameters of the models and correspondingly strawberry yield was considered as output variable. Coefficient of determination (R-2), root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) were calculated as 0.9121, 0.0667, 0.0388 and 0.0084 for ANN model and they were calculated as 0.963, 0.017, 0.014 and 0.003 for ANFIS model. Finally, a comparison between developed ANN and proposed ANFIS was made and the outcomes disclosed that ANFIS model can predict strawberry yield relatively better than does ANN model. (C) 2013 Elsevier Ltd. All rights reserved.