Prediksi Harga Emas Menggunakan Metode Generalized Regression Neural Network Dan Algoritma Genetika

Marthasari, Gita Indah (2014) Prediksi Harga Emas Menggunakan Metode Generalized Regression Neural Network Dan Algoritma Genetika. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 5111202002_Master Thesis.pdf - Accepted Version Download (2MB)

Abstract

Prediksi harga emas merupakan aktivitas penting bagi banyak pihak. Salah satu metode prediksi harga emas yang dapat digunakan adalah jaringan syaraf tiruan berbasis generalized regression neural network (GRNN). Dalam penelitian sebelumnya, GRNN digabungkan dengan teknik dekomposisi Seasonal Trend Decomposition based on Locally Weighted Regression (STL) dan metode theta. Kinerja GRNN dipengaruhi oleh data latih yang digunakan karena ukuran jaringan yang terbentuk akan berbanding lurus dengan jumlah data latih. Untuk mengatasi meningkatnya ukuran jaringan seiring dengan bertambahnya data latih, proses reduksi data latih tanpa mengurangi akurasi prediksi perlu dilakukan. Dalam penelitian ini, metode peramalan GRNN diintegrasikan dengan algoritma genetika untuk mereduksi data latih guna menghasilkan model peramalan yang lebih efisien. Sebelum diramalkan, data harga emas didekomposisi menggunakan STL menjadi komponen data musiman, tren, dan residual. Ketiga komponen data tersebut diramalkan menggunakan dua metode yang berbeda, yaitu GRNN untuk meramalkan komponen data musiman dan residual, dan metode theta untuk meramalkan komponen data tren. Hasil peramalan dari ketiga tersebut selanjutnya digabungkan menggunakan jaringan syaraf tiruan propagasi balik untuk memperoleh hasil peramalan akhir. Hasil pengujian menunjukkan bahwa GRNN yang diintegrasikan dengan algoritma genetika, selain mampu menghasilkan peramalan dengan akurasi yang setara dengan GRNN tanpa algoritma genetika, juga mampu memberikan akurasi yang lebih baik dibandingkan dengan hasil permalan menggunakan model peramalan Arima. Selain itu, kombinasi GRNN dengan algoritma genetika mampu mereduksi jumlah data latih sebesar 50% dan mampu mengurangi waktu proses peramalan sebesar 34%
==============================================================================================================================
The prediction of gold price is an important activity for all parties. One of the gold prediction methods that can be used is the artificial neural network based on the generalized regression neural network (GRNN). In the previous research, GRNN was combined with the decomposition technique of Seasonal Trend Decomposition based on locally weighted regression (STL) and the Theta method. The GRNN performance was influenced by thetraining data size used,since the network size formed is proportionally dependent on the training data size. To cope with the increasing of the network size along with the increasing of the training data size, the process of training data reduction that is capable of maintaining the accuracy of the prediction is interesting to investigate. In this research, the GRNN prediction method is integrated with the genetic algorithm in order to reduce the training data sizethat will in turn produce the more efficient prediction model. Initially, the gold price data is decomposed into three components; i.e., seasonal, trend, and residual data using STL decomposition technique. Those three components are then predicted using two different methods, namely GRNN to predict the component of the seasonal and residual data, and the theta method to predict the trend data component. The prediction results of those three components are combined together using the back propagation neural network algorithm to obtain final results of the prediction. Experimental results showed that the GRNN method that was integrated with the genetic algorithm was not only capable of producing prediction results with the accuracy similar to those produced using the original GRNN method, but also capable of giving the better prediction accuracy in compared to that produced using ARIMA prediction model. In addition to that, the combination of GRNN with the genetic algorithm is also capable of reducing the amount of the training data as much as 50% and reducing the computing time consumed by the process as much as 34%

Item Type:	Thesis (Masters)
Additional Information:	RTIf 005.1 Mar p
Uncontrolled Keywords:	peramalan harga emas, optimasi data latih, algoritma genetika, dekomposisi data runut waktu, generalized regression neural network, metode theta
Subjects:	Q Science > QA Mathematics > QA402.5 Genetic algorithms.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Mr. Tondo Indra Nyata
Date Deposited:	23 Jun 2023 07:00
Last Modified:	23 Jun 2023 07:00
URI:	http://repository.its.ac.id/id/eprint/98214

Actions (login required)

View Item