PREDIKSI CURAH HUJAN HARIAN DI KABUPATEN GUNUNGKIDUL PROVINSI DAERAH ISTIMEWA YOGYAKARTA DENGAN JARINGAN SYARAF TIRUAN METODE BACKPROPAGATION; DAILY RAINFALL PREDICTION IN GUNUNGKIDUL REGENCY SPECIAL PROVINCE OF YOGYAKARTA USING BACKPROPAGATION NEURAL NETWORK

Darnanto, Anifuddin Azis

2015 | Skripsi | FMIPA

Gunungkidul Regency is region in Indonesia are vulnerable to the negative impacts of climate change, especially drought and rainfall anomalies. Natural relief factor in Gunungkidul predominantly limestone hill aggravate drought in Gunung Kidul. The impact of drought effect on the socio-economic conditions of people in Gunung Kidul Regency who work as farmers. Daily rainfall prediction is one efforts that can be done to mitigate the impact of drought in Gunungkidul. Artificial neural networks (ANN) method of backpropagation is a method that can be used for predictive purposes. ANN is suitable for the purpose of prediction because ANN able to study the data patterns in the past so as to produce a prediction formula based on the difference between the target with a network output. In this research the implementation of backpropagation neural network(BPNN) method to look for daily rainfall prediction model in Gunungkidul. Network training using 205 datasets and network testing using 88 datasets. Maximum accuracy is 78% of network training and network testing maximal accuracy is 68%. Parameters trainers who produce the best accuracy are the number of hidden nodes 400-500 units, the momentum factor of 0,1, and learning rate of 0,05.

Kata Kunci : N