The cryptocurrency market has become one of the fastest-growing sectors in the world, but its extreme price swings often make investors cautious. To address this, researchers from Universitas Padjadjaran developed a machine learning-based forecasting model that can predict cryptocurrency price movements with high accuracy.
Using the Gated Recurrent Units (GRU) algorithm, the study focused on historical Bitcoin price data from January 2017 to July 2024. Unlike previous research that relied only on accuracy, this study also considered the model’s ability to “fit” the data pattern effectively. The combination of accuracy and learning curve analysis made the results more reliable.
The model achieved impressive performance, with a Mean Absolute Percentage Error (MAPE) of just 1.95%, indicating very precise predictions. By applying the sliding window technique, the model could quickly learn time series patterns, making it both efficient and accurate.
This research contributes to SDG 8: Decent Work and Economic Growth, as it supports financial innovation and stability in the digital economy. With more accurate forecasting tools, investors and financial institutions can make better decisions, reduce risks, and contribute to sustainable economic growth.
#UnpadResearch #Cryptocurrency #MachineLearning #SDG8
Link to the paper: https://www.scopus.com/pages/publications/105002086503
26/Mat/2025
