Abstract: This research presents a study on the development of a model for oil palm yield using neural network approach. The structure of this neural network requires the identification of the input variables and the output. We identified that the percentages of nitrogen, phosphorus, potassium, calcium and magnesium in leave were used as input variables and fresh fruit bunch was used as the target variable. An investigation of the combinations of activation function in the input layer to the hidden layer and the hidden layer to the output layer found that each combination also affects the neural network performance. The effect of the learning rate, momentum term, number of runs and number of hidden nodes was also investigated. The number of hidden nodes was found to significantly affect the neural network performance. However, the learning rate, momentum term and number of runs were found to have an insignificant effect on the neural network performance. Using R2 values the suitability of the models were measured. Results demonstrate that the neural network model outperformed regression analysis, which can be considered as alternative in modeling of oil palm yield.