<![CDATA[Cryptocurrency is one of the digital assets that is increasingly popular for investment in Indonesia. However, the price movements of cryptocurrencies tend to be volatile, as prices can change at any time and are not easy to predict. This study aims to predict cryptocurrency price movements using the Long Short-Term Memory Algorithm (LSTM) and Moving Average Convergence Divergence (MACD). LSTM is an algorithm used to generate optimal weights and biases in modeling cryptocurrency data, while MACD is used to analyze trends and momentum in cryptocurrency prices. The data used consists of daily closing prices of Bitcoin (BTC), totaling 809 data points. The data is divided into 70% (566 data) for the training process and 30% (243 data) for the testing process. From this data, patterns are formed with five inputs and one output, resulting in 561 patterns for the training process and 238 patterns for the testing process. The LSTM and MACD processes for predicting cryptocurrency include procedures for data input, data division, parameter initialization, LSTM calculation, average error evaluation, and MACD calculation. Based on the program implementation, with several parameter values, the average error difference obtained during the training stage is 0.0695 and 0.0303 during the testing stage. Because the average error difference obtained is relatively small, this indicates that LSTM-MACD is capable of recognizing data patterns and predicting data effectively.]]>
Copyrights © 2026
