JISARA

Abstract: The volatility and unpredictability of cryptocurrency lead to financial losses for investors. We develop a predictive portfolio mobile app called CryptoProphet that leverages deep learning models to predict future prices and help crypto traders make informed decisions. A unique approach called the Individualized Model Selection (IMS) Strategy is adopted instead of relying on an ensemble or single model type across all cryptocurrencies. The IMS Strategy involves training each of the 30 cryptocurrencies using Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSRM (Bi-LSTM) models. Then, the best-performing model for each cryptocurrency is selected for next-day price predictions using performance metrics of Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R2). This research addresses the highly volatile nature of cryptocurrencies for ensuring accurate predictions. This approach includes collecting historical data, preprocessing it, and training the models on sequences of price data. The evaluation of the models using the aforementioned metrics confirms their effectiveness. The app seamlessly integrates these predictions, providing users real-time price forecasts and essential market insights. The findings showed that the CryptoProphet portfolio app predicts prices accurately, reducing risks and maximizing profits in the volatile cryptocurrency market. Future work will focus on improving prediction accuracy by incorporating sentiment analysis and additional features such as market capitalization and volume to further improve prediction accuracy.

Download this article: JISARA - V18 N3 Page 4.pdf

Recommended Citation: Shewarade, Y., Aly, S., Yoon, H., Chung, S., (2025). CryptoProphet: Building a Cryptocurrency Portfolio App with Integrated Market Predictive Models. Journal of Information Systems Applied Research and Analytics 18(3) pp 4-17. https://doi.org/10.62273/NNLD2781