Deas Achmad Rivai
Teknik Informatika STMIK ATMA LUHUR PANGKALPINANG
Jl. Jend. Sudirman Selindung Lama Pangkalpinang Kepulauan Babel
email : deasrivai@gmail.com
Abstrak
Daily weather forecasts method developed by BMKG currently are subjective, or in other words are still very dependent on the operator. This study aims to develop a method of daily weather forecasts are objective. Objective means is by including certain data it will automatically obtained value forecasts, so no more subjective elements of the forecaster. Making an application is using the waterfall method which refers to the rules of Classic Life Cycle, which step by step through having to wait the completion of the previous stage. Methodology software which is used in the form of data gathering phase of the literature study and observation. Applications used in the manufacture of this application is version 7.7 matlab R2008b and degrib. Comparison between the output and the target backpropagation training produces value R = 0.99975, the results of the test with the value R = 0.7462, maximum error = 28.6841, and the minimum error = 0. Comparison between the output and the target training LVQ generate value R = 0.6305. Of the correlation value obtained training and testing, the network is fit for use for the next day’s forecast rainfall. Based on the results of the correlation test, weather parameters that determine rainfall in Pangkalpinang is rh 700, 700 spfh, rh 500, rh 850 (air humidity layer 850, 700, and 500 mb) and ugrd-10 (U component of wind at a height of 10 meters). These parameters are used as input applications. Advice given to the results obtained allow better is the use of a nearest grid points around the study site, increase the length of the data used, and try to use another network with a different algorithm.
Kata Kunci : Weather Forecast, Artificial Neural Network
