Assessment of machinery energy ratio in potato production by means of artificial neural network

A single hidden layer Artificial Neural Network (ANN) model was developed to estimate a machinery energy ratio (MER) indicator, used to characterize and assess mechanization status of potato farms in Iran with a view point of energy expenditure in farm machinery. A wide range of variables of farming activities were examined. Initially, 90 attributes were used as input variables to predict desired MER output. Using regression analysis, 13 inputs were finally selected to model MER. Performance of developed ANN model was evaluated with various statistical measures including the coefficient of determination (R2), mean absolute percentage error (MAPE), mean squared error (MSE) and mean absolute error (MAE). The optimum ANN model had a 13 - 4 - 1 configuration. The values of the optimum model’s outputs correlated well, with R2 of 0.98. Value of MAPE calculated as 0.0001 for best ANN model, which indicate superiority of this model over other prediction models. Sensitivity analyses were also conducted to investigate the effects of each input item on the output value. Since the ANN model can predict this mechanization indicator for a target farming system in Hamadan province of Iran, it could be a good estimator for appraising mechanization of other regional farms. Also it overcomes some of the limitations of using simple data available from local databases as inputs that may contain errors.