Pwenerapan JST(Backpropagation) untuk Prediksi curah hujan (Studi kasus: Kota Pekanbaru)

Abstract

Weather forecasting is an important thing in our lives and can help us to minimize the impact it will have in the future, we need a very high accuracy in forecasting the weather conditions in the future. For that, he built a system that can predict the weather. In this study using an artificial neural network models are often used in forecasting. The method used is Backpropagation where there is a training process the data (making patterns), forecasting and output in the form of a state system of rainfall and its weight. In this study using binary sigmoid activation function (logsig) and bipolar sigmoid (tansig) using parameters epoch 1000, learning rate 0.01 and error 0.001. The data used from the city of Pekanbaru BMKG station with focus testing system for local area Pekanbaru. Output adjusted into 5 categories: bright, light rain, moderate rain, heavy rain and torrential rain by BMKG standards. An accuracy system is 96 %, where most of the failures are in the category of moderate rain and heavy rain.
Keywords: Backpropagation, BMKG, Pekanbaru, Weather Forecast