Investasi sebagai komitmen sumber daya dengan ekspektasi keuntungan di masa depan, mengalami pergeseran fokus akibat resesi ekonomi. Melemahnya nilai mata uang mendorong investor melirik instrumen rendah risiko seperti emas. Dalam situasi ini, prediksi harga emas menjadi krusial bagi pengambilan keputusan investasi. Namun, kompleksitas faktor yang memengaruhi fluktuasi harga emas, seperti inflasi, nilai tukar mata uang, suku bunga, hingga kondisi geopolitik, menyulitkan prediksi akurat hanya dengan metode statistik tradisional. Penelitian ini mengkaji penerapan model Long Short-Term Memory (LSTM), sebuah algoritma deep learning yang dikenal tangguh dalam mempelajari pola data time-series untuk memprediksi harga emas dunia. Data historis harga emas harian periode 01 Januari 2020 hingga 01 Januari 2024 digunakan dalam pengembangan dan pengujian model. Penelitian ini secara sistematis mengeksplorasi berbagai konfigurasi model LSTM, dengan mengoptimasi parameter menggunakan algoritma Adaptive Moment Estimation (ADAM), untuk mencapai akurasi prediksi terbaik. Kinerja model dievaluasi menggunakan metrik Root Mean Squared Error (RMSE). Hasil penelitian menunjukkan model LSTM dengan arsitektur 8 lapis hidden layer yang terdiri dari 1, 32, dan 64 neuron, serta dilatih dengan epoch 1, 2, dan 3, mencapai performa optimal dengan nilai RMSE sebesar 22.17. Model terbaik ini kemudian diuji pada data uji dan menghasilkan nilai RMSE sebesar 3.82361, menunjukkan kemampuan prediksi yang sangat baik dan potensi penerapannya dalam skenario investasi emas.

Investment as a commitment of resources with the expectation of future returns, experienced a shift in focus due to the economic recession. In this situation, gold price prediction becomes crucial for investment decision-making investment decision-making. However, the complexity of factors that influence gold price fluctuations, such as inflation, currency exchange rates, interest rates, to geopolitical conditions, makes it difficult to make accurate predictions using only traditional statistical methods traditional statistical methods. This research examines the application of the Long Short-Term Memory (LSTM) model, a deep learning algorithm known to be robust in algorithm known for its robustness in learning time-series data patterns for predicting the world gold price. Historical data of daily gold prices for the period of 01 January 2020 to 01 January 2024 is used in the development and testing of the model development and testing. This research systematically explores various configurations of LSTM models, by optimizing parameters using the Adaptive Moment Estimation (ADAM) algorithm, to achieve the best prediction accuracy. The performance of model is evaluated using the Root Mean Squared Error (RMSE) metric. The results show that the LSTM model with an architecture of 8 layers of hidden layer architecture consisting of 1, 32, and 64 neurons, and trained with epochs of 1, 2, and 3, achieved optimal performance with an RMSE value of 22.17. The best model This best model was then tested on test data and produced an RMSE value of 3.82361, demonstrating its excellent predictive ability and potential for applicability in gold investment scenarios.