At present, with the rapid development of blockchain technology and digital currencies, digital assets have gradually become a crucial component of investment portfolios, with Bitcoin standing out as the most prominent. Our aim is to reduce the investment risk for stakeholders in the highly volatile Bitcoin market by expanding predictions beyond Bitcoin and into other digital currencies. Through the application of machine learning models tailored for time series data and deep learning models, we have discovered methods for predicting Bitcoin prices with a high degree of accuracy. This article centers around predicting the closing price of Bitcoin using attributes from various categories of digital currencies, including Bitcoin itself, Ethereum and stablecoins. We employ multiple evaluation techniques to assess the performance of our models. Our findings indicate that using Bitcoin’s attributes or stablecoins pegged to Bitcoin can achieve accuracy rates of up to 98%, with Ethereum closely following, boasting accuracy rates exceeding 90%. In the practical application domain, we innovatively employed a stablecoin model pegged to Bitcoin for backtesting Bitcoin returns. This model not only filled a significant gap in this field but also surpassed models trained solely on Bitcoin itself, yielding returns that exceeded 40 times. This approach underscores the potential of diversifying predictive analytics in digital asset investments.