"STUDY OF LINE SEARCH TECHNIQUES ON THE MODIFIED BACKPROPAGATION FOR FORECASTING OF WEATHER DATA IN INDONESIA"

“STUDY OF LINE SEARCH TECHNIQUES ON THE MODIFIED BACKPROPAGATION FOR FORECASTING OF WEATHER DATA IN INDONESIA”

Author		: UNTARI NOVIA WISESTY; FHIRA NHITA
Published on	: Far East Journal of Mathematical Sciences (India)

Abstract

The extreme alteration of weather condition in Indonesia at the last two decades is the main idea that encourages this research. Recently, we have been built an improvement system of back propagation performance by using conjugate gradient on forecasting of air temperature and humidity in Indonesia [1]. Data of weather used in this research is derived from Indonesian Agency for Meteorology Climatology and Geophysics covering air temperature, rainfall, air humidity, length of sun radiation, air pressure, and wind velocity in Kemayoran Jakarta area along five years (2007-2012). In this paper, we examine a technique of line search on conjugate gradient for backpropagation training in the forecasting of temperature, humidity, and air pressure. The technique of line search plays a part in determining step size parameter on conjugate gradient to minimize the process of searching. Eventually, the proposed method provides good performance, in addition to the training needs of small epoch compared to the standard backpropagation.