Published Paper Details:

The cryptocurrency industry, led by Bitcoin, has grown in popularity and significance in the past few years, capturing the interest of shareholders, buyers and sellers, and experts alike. Forecasting the future value of Bitcoin presents a formidable challenge yet holds substantial importance. In this project, we harness the capabilities of two potent deep learning algorithms: Long Short Term Memory (LSTM) and eXtreme Gradient Boosting (XGBoost) to forecast the value of bitcoin. LSTM, a deep learning model recognized for its aptitude in capturing sequential dependencies, is paired with XGBoost, an ensemble method adept at handling intricate, non-linear relationships. Our approach revolves around the analysis of historical Bitcoin price data, employing these two algorithms in tandem to furnish more precise and robust predictions. The overarching goal of this endeavor is to enrich our understanding of the dynamics within the Bitcoin market and to offer valuable insights to cryptocurrency enthusiasts and investors. Bitcoin, as the foremost digital currency, has garnered substantial attention from investors, leading to exponential growth. Nevertheless, the task of predicting Bitcoin's price remains daunting due to the inherent unpredictability of the market. Initially, our study intended to incorporate sentiment analysis, but complications related to data mapping led us to rely exclusively on historical Bitcoin transaction data. This alternative approach has yielded a notably accurate model, offering invaluable insights for prospective investors. Our investigation delved into various recurrent neural network (RNN) structures, encompassing LSTM, subjecting them to exhaustive testing, which entailed fine-tuning the number of activation function layers and units. In this comparative analysis, LSTM emerged as the leading performer in sentiment analysis among the predictive models. Given the exceptional volatility of Bitcoin, with price fluctuations approximately six times more pronounced than those of traditional fiat currencies, it is imperative to employ a model capable of handling long-term time series dependencies, a realm where LSTM excels.