<![CDATA[Over the past few years, the implementation of renewable energy or go-green has intensified along with the rapid development of its technology and increasing uncertainty of natural conditions that cause the prices of non-ferrous metals such as copper, aluminum, nickel, etc. used as main components for developing renewable energy devices, e.g.: battery, experience instability price in the commodity futures market. Economic players who trade metals in the futures market certainly need to be careful and must evaluate the state of the world economy. This study proposes a prediction engine as a combination of Bidirectional Long-Short Term Memory (BiLSTM), with three optimization algorithms, i.e.: Adam, Root Mean Squared Propagation (RMSProp), and Stochastic Gradient Descent (SGD), and transfer learning to make model training better. Experiments on four historical data on nickel, lead, aluminum and copper prices in the commodity futures market are conducted. The selected features are: open price, close price and volume price. Twelve models will be created to find the model that best predicts the metal prices. The top 3 models with the best performance were selected, they are: model 4 RMSProp with R2 value of 0,99029 and MSE 0,00076 as the first ranking, model 3 Adam with R2 value of 0,98877 and MSE 0,00074 as the second ranking, and model 4 Adam with value of R2 0,98522 and MSE 0,00115 as the third ranking.]]>
Copyrights © 2024
