STATISTICAL TECHNIQUE DAN PARAMETER OPTIMIZATION PADA NEURAL NETWORK UNTUK FORECASTING HARGA EMAS

Harminto Mulyo

Abstract


ABSTRACT
Gold is a precious metal that is valuable in the world that is soft, corrosion-resistant, malleable. The investment experts often advise to invest in gold because gold is a classic hedge against inflation and adds value in conditions of instability of currency exchange rate fluctuations. Gold price history data from year to year tends to rise, it's interesting researchers to examine the data using a variety of forecasting methods, including Statistical Technique and Data Mining Technique. Experiments were performed to look for the smallest error value by using statistical techniques and data mining. From the test results, the best statistical technique that uses a technique Single Moving Average with MSE of 760.55. In the data mining technique using Neural Network Backprogration obtained RMSE of +/- 22 730 6945.
Keywords: Parameter, Optimization, gold, Neural Network
ABSTRAK
Emas merupakan salah satu logam mulia yang bernilai di dunia yang bersifat lunak, tahan korosi, mudah ditempa. Para pakar investasi seringkali menganjurkan untuk berinvestasi pada emas karena emas merupakan sarana lindung nilai klasik untuk melawan inflasi dan menambah nilai dalam kondisi ketidakstabilan fluktuasi nilai mata uang. Data riwayat harga emas dari tahun ke tahun cenderung naik, hal tersebut menarik peneliti untuk menguji data menggunakan berbagai metode peramalan, diantaranya Statistical Technique dan Data Mining Technique.Eksperimen dilakukan untuk mencari nilai error terkecil dengan menggunakan teknik statistik dan data mining. Dari hasil pengujian, teknik statistik terbaik yaitu menggunakan teknik Single Moving Average dengan MSE sebesar 760.55. Pada teknik data mining dengan menggunakan metode Neural Network Backprogration didapat RMSE sebesar 22.730 +/- 6.945.
Kata Kunci: Parameter, Optimization, emas, Neural Network

Full Text:

FULL TEXT (PDF)


DOI: https://doi.org/10.34001/jdpt.v7i2.432

Article Metrics

Abstract view : 316 times
FULL TEXT (PDF) - 356 times



Disprotek Indexed by:

1 Google Scholar  2 BASe3 Onsesearch 4 Garuda 5 Sinta 6 Dimensions7 Crossref 8 JurnalStories 9 ROAD 10 ICE11 ORCID  

Visitor Statistics
Web
Analytics Made Easy - StatCounter
Flag Counter

Lisensi Creative Commons

DISPROTEK: Journal of Informatics Engineering, Information Systems, Electrical Engineering, Industrial Engineering, Civil Engineering, and Aquaculture is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.