Penggunaan Elman Recurrent Neural Network dalam Peramalan Suhu Udara sebagai Faktor yang Mempengaruhi Kebakaran Hutan

Abstract

Recent forest and land wildfires in Indonesia not only cause to the ecosystem lost but also economical lost as well as health and polution effect. On the other hand, fire weather is concerned to be an important aspects for fire occurrences. Fire weather directly affects fuel temperature, which accelerating its easiness to be burnt. Thus, it is important to clarify the effects of air temperature to forest fire incidence. A statistical analysis and forecasting for air temperature is used to predict the future air temperature condition. In this research we use Elman Recurrent Neural Network (ERNN) to predict the temperature for a few days ahead, then the result of this prediction will be compared by using ARIMA (Ramdani, 2011). In this research we use the air temperature data of 2001- 2004 which implemented using MATLAB. The best result in this research is one day ahead prediction with RMSE 0.51 and MAPE 1.55%. ERNN gave the better performance with MAPE (ERNN) 1.55% and MAPE (ARIMA) 3.11% (Ramdani, 2011).