APLIKASI PREDIKSI PRODUKSI PADI MENGGUNAKAN REGRESI INTERVAL DENGAN NEURAL FUZZY DI KABUPATEN KUBU RAYA
Abstract: Paddy (oryza sativa)
was a basic foodstuff that was vital for people of Indonesia. The amount of
rice production at a time can not be calculated exactly, so it was necessary to
predict rice production in order to provided rice to achieve endurance and food
self-sufficiency. This research aimed to make prediction system of rice
production in january-april period,
may-august period, and september-december period using interval regression with
neuro fuzzy in Kubu Raya regency. The interval regression method was based on
backpropagation network which there are two separately trained backproagation
networks, one model to find out the lower bound and one model to find out the
upper bound. Based on the results of the training and testing phase, the best
network result were obtained by 4 input layer neurons, 3 hidden layer neurons
and 1 output layer neuron using minimum error parameter 0,00001 and maximum
epoch 100000. The results showed that MSE BPN- for period 1 training with lr
0.09 of 0,064722 and MSE BPN+ of 0,030644; Period 2 with Ir 0.09 results of MSE
BPN- 0,141674 and MSE BPN+ of 0,179612, and period 3 with Ir 0.09 results of
MSE BPN- of 0,025324 and MSE BPN+ of 0,036961. The mean deviation of the
interval between the lower limit with the actual value on the test data was
4.369,2 (BPN-), while the interval deviation between the upper limit with the
actual value on the test data was 19.744,3 (BPN+).
Penulis: Kurniati Asih , Fatma
Agus Setyaningsih , Dwi Marisa Midyanti
Kode Jurnal: jptkomputerdd170030
