PREDIKSI HARGA BITCOIN BERDASARKAN DATA HISTORIS MENGGUNAKAN MODEL HYBRID TEMPORAL CONVOLUTIONAL NETWORK, BIDIRECTIONAL LONG SHORT - TERM MEMORY, DAN GATED RECURRENT UNIT

Bitcoin has become a digital asset with extremely high price volatility, making it an attractive object for both investment and research. This phenomenon drives the need for predictive models capable of capturing the temporal complexity and non
linearity of the crypto market data. This research aims to develop and test the performance of a hybrid model that combines a Temporal Convolutional Network (TCN), Bidirectional Long Short-Term Memory (BiLSTM), and Gated Recurrent Unit (GRU) to predict the closing price of Bitcoin. This model is designed to
leverage the advantages of each architecture, where TCN functions as a temporal feature extractor, BiLSTM processes data from two directions to capture long-term dependencies, and GRU filters relevant information for the final prediction. This
study uses a quantitative method with historical daily data of Bitcoin's closing prices from January 1, 2018, to March 31, 2025, obtained from the investing.com website. The data was divided into 80% training data and 20% testing data. The
evaluation results on the test data show that the TCN-BiLSTM-GRU hybrid model performs exceptionally well, achieving a Mean Absolute Percentage Error (MAPE) of 5.96% and a Root Mean Squared Error (RMSE) of 7,757.55. The low MAPE value demonstrates the model's ability to follow market trends and volatility with good accuracy. The developed model was then implemented into an interactive web application using Streamlit, which is capable of presenting Bitcoin price predictions in real-time.
Keywords: Price Prediction, Bitcoin, Hybrid, Temporal Convolutional Network (TCN), Bidirectional Long Short-Term Memory (BILSTM), Gated Recurrent Unit (GRU), Historical Data.