PENDEKATAN DEEP LEARNING MENGGUNAKAN METODE LSTM UNTUK PREDIKSI HARGA BITCOIN

Abstract

Bitcoin price prediction involves analyzing a variety of factors, including market sentiment, trading volume, economic news, technological developments, and other factors that affect supply and demand. Both technical and fundamental analysis methods can be used to try to predict Bitcoin price movements. In this Bitcoin price prediction using a Deep Learning approach with the chosen method is LSTM. The LSTM (Long Short-Term Memory) method is a popular type of Recurrent Neural Network (RNN) model for predicting the price of Bitcoin and other financial assets. LSTM can solve the problem of price movements that have long-term dependencies, which traditional RNN models cannot handle well. LSTMs have the ability to "remember" information from longer periods of time, thereby recognizing complex patterns and trends in historical data. In this study the prediction period used a dataset from March 1 2016 to November 24 2018. This study used an epoch parameter of 10 with a learning rate of 0.001. In addition, the batch size parameter used is 25 with layers only. The evaluation results of this study resulted in an RMSE of 77.74 and an MAE of 278.33. This shows that the RMSE value is small because the Bitcoin price range is too far.