Prastanti, Novianti Mega (2015) Peramalan harga emas yang dipengaruhi oleh minyak, nilai tukar mata uang; dan indeks standard dan poor menggunakan metode general regression neural network. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
5211100002-Undergraduate-Thesis.pdf Download (3MB) | Preview |
Abstract
Emas merupakan salah satu aset investasi yang paling diminati
karena bersifat mudah diuangkan, sehingga emas menjadi
salah satu jenis investasi yang menarik. Dalam beberapa
penelitian sebelumnya ditemukan bahwa fluktuasi harga emas
dipengaruhi oleh tiga variabel penting, yaitu harga minyak,
nilai tukar mata uang, dan indeks Standard & Poor (SP). Oleh
karena fluktuasi harga emas yang tidak menentu dapat
menyulitkan para pelaku bisnis dalam menentukan waktu yang
tepat untuk membeli atau menjual emas, maka model peramalan
yang memperhitungkan pengaruh dari ketiga variabel
penting tersebut menjadi esensial untuk disediakan.
Dalam Tugas Akhir ini, untuk memperoleh model peramalan
yang akurat dalam memperhitungkan pengaruh variabel harga
minyak, nilai tukar mata uang, dan indeks SP, dikembangkan
sebuah model peramalan yang mengintegrasikan metode
peramalan berbasis Generalized Regression Neural Network
(GRNN) yang melibatkan teknik dekomposisi Seasonal Trend
x
Decomposition based on Loess (STL) dan metode Theta. Teknik
dekomposisi STL digunakan untuk memisahkan komponen data
musiman (seasonal), data tren, dan data residual (irregular).
Metode peramalan berbasis GRNN kemudian digunakan untuk
meramalkan komponen data musiman dan data residual. Di
lain pihak, metode Theta digunakan untuk meramalkan komponen
data tren. Kedua hasil peramalan ini kemudian diintegrasikan
menggunakan Backpropagation Neural Network
(BPNN) untuk memperoleh hasil peramalan akhir.
Model peramalan yang dikembangkan dalam Tugas Akhir ini
di uji coba menggunakan data sebanyak 2.484 data harian yang
diperoleh dari berbagai sumber untuk kurun waktu mulai 3
Januari 2005 hingga 31 Desember 2014. Dengan menggunakan
model peramalan yang dilatih dengan menggunakan
data untuk periode sembilan tahun pertama memberikan hasil
Mean Absolute Percentage Error (MAPE) sebesar 0,60% untuk
uji coba validasi yang melibatkan data satu tahun terakhir.
Selain itu, hasil uji coba menunjukkan bahwa model peramalan
terbaik diperoleh apabila dilakukan pembentukan model baru
untuk setiap kali terdapat penambahan data aktual selama satu
bulan terakhir. Dan hasil uji coba pengaruh kombinasi variabel
mulai dari variabel tunggal, ganda, dan triplet menunjukkan
bahwa harga minyak mempunyai pengaruh yang cukup
signifikan terhadap peramalan harga emas.
==================================================================================================
Gold is one of the most interested investment asset because it is
easily cashed, and hence gold became one of the attractive
investments. In some previous studies found that fluctuations in
the price of gold could be influenced by three important variables,
namely oil prices, exchange rates, and the Standard &
Poor's (SP) indexes. Since uncertain fluctuations in the price of
gold may complicate the businessmen in determining the right
time to buy or sell gold, then the forecasting model that takes
into account the influence of those three important variables
become essential to be provided.
In this final project, to obtain an accurate forecasting model
that takes into account the influence of oil prices, exchange
rates, and SP indexes, a forecasting model that integrates forecasting
method based on Generalized Regression Neural Network
(GRNN) which involves the decomposition technique Seasonal
Trend Decomposition based on Loess (STL) and Theta
method is developed. The STL decomposition technique is used
xii
to separate components of the seasonal data, trend data, and
the irregular data. The GRNN-based forecasting method is then
used to predict seasonal and residual data components. On the
other hand, the Theta method is employed to predict trend data
component. Finally, both of these forecasting results are integrated
using a Backpropagation Neural Network (BPNN) to obtain
the final forecasting results.
Forecasting model that has been developed in this final project,
tested using 2,484 daily data ranging from January 3, 2005 until
December 31, 2014 obtained from various sources. The first
nine-year period of data was used to build the forecasting
model, while the last year data was used to validate the forecasting
results. A very accurate results of 0.6% for their mean
absolute percentage error (MAPE) were obtained. In addition
to this, experimental results also showed that the forecasting
model must be rebuilt in accordance with the availability of recent
actual monthly data. Moreover, among the combinations
of three variables being involved in the forecasting model, oil
prices become the most significant one that influences the gold
price forecasting.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Additional Information: | RSSI 005.1 Pra p |
| Uncontrolled Keywords: | peramalan, generalized regression neural network, harga emas, harga minyak, nilai tukar mata uang, indeks SP, Seasonal Trend Decomposition based on Loess, metode theta, jaringan syaraf tiruan propagasi balik |
| Subjects: | Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) |
| Divisions: | Faculty of Information Technology > Information System > 57201-(S1) Undergraduate Thesis |
| Depositing User: | Mr. Marsudiyana - |
| Date Deposited: | 15 Nov 2019 02:18 |
| Last Modified: | 15 Nov 2019 02:18 |
| URI: | http://repository.its.ac.id/id/eprint/71807 |
Actions (login required)
 |
View Item |