Analysis of ANTAM Gold Prediction Using Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) Algorithm Nur Aziza Luxfiati (a*), Alhadi Bustamam(a), Titin Siswantining (a)
(a*) Department of Mathematics, Universitas Indonesia, Depok, Indonesia
*nur.aziza11[at]ui.ac.id
Abstract
Gold is an important component in banking and stock markets^ economic and financial status. Gold has a fluctuating price trend. Various factors influence gold price fluctuations, including economic factors such as the supply and demand of gold. It is therefore difficult to accurately predict trends with traditional statistical models. This research uses Antam^s gold stock price as data and compares the LSTM and GRU algorithms in predicting gold prices. Therefore it is hoped to obtain an algorithm with the best level of accuracy as a consideration for investors in making investment decisions. The results of this research state that predicting gold prices using the GRU algorithm is better than LSTM. The GRU model using batch size 32 and epoch 150 obtains an accuracy of 98,26% while the LSTM model using batch size 32 and epoch 150 obtains an accuracy of 96,26%.
Keywords: Gold, Using Long Short-Term Memory, Gated Recurrent Unit, Prediction
