<![CDATA[Citrus is one of the horticultural plants which are popular in Indonesia but the citrus production from year to year has fluctuated. There are some main causes that affected to the fluctuation of national production of citrus which are climate, environment, and diseases. One way to overcome the climate, environment, and diseases of citrus production is to provide fertilizer at the right dose and proportional to that matched with the environment and its characteristics. This study aims to forecast the dosage of citrus fertilizer according to the characteristics and environment. This study uses Artificial Neural Network (ANN) backpropagation. The architecture a network of 3 nueron input layer that represents the related parameters is width of the canopy, soil texture and rainfall, one hidden layer, and 3 nueron output layer that represents the composition of the fertilizer that is nitrogen, phosphorus, and potassium. The best network architecture design for forecasting doses of citrus fertilizer are 3 input neurons, 5 nueron hidden layer and 3 output neurons. The value of the learning rate used is 0.3 with the maximum iteration of 500 and the training data is 56 and the test data 8. The Mean Absolute Precentage Error (MAPE) evaluation value of the composition data of the fertilizer dose is 9.178% obtained from average error of dose of nitrogen, phosphorus, and potassium fertilizer.]]>
Copyrights © 2017
