<![CDATA[This study employs a Long Short-Term Memory (LSTM) model to forecast the daily closing price of PT Pertamina Geothermal Energy Tbk (PGEO) and to assess its training and out-of-sample performance. The dataset comprises 233 daily closing-price observations from April 8, 2025, to March 31, 2026. The data were split chronologically into 186 training observations and 47 testing observations. Min-Max scaling and a 60-day look-back window were used to construct the input sequences. The model consisted of two LSTM layers with 64 and 32 units, each followed by a dropout layer with a rate of 0.20. It was trained for 100 epochs with a batch size of 32 using the Adam optimizer and mean squared error as the loss function. Performance was evaluated separately using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The training data produced an MSE of 3,322.34, an RMSE of 57.64, and a MAPE of 3.10%, while the testing data produced an MSE of 4,230.63, an RMSE of 65.04, and a MAPE of 5.02%. The higher testing errors indicate a moderate decline in performance on new observations. The model followed the broad price movement but had difficulty capturing abrupt daily changes. A recursive five-step forecast generated predicted prices ranging from IDR 951.32 to IDR 968.81. This increase applies only to the specified forecast horizon and should not be interpreted as evidence of a continuing upward trend or as investment advice.]]>
Copyrights © 2026
