ADS Capstone Chronicles Revised

21

8.0.4 Limitations

transparently, instills confidence in the model ’ s reliability.

Several limitations were encountered during this study. Firstly, the granularity of the data was limited to daily observations, which needs to be improved to capture the high-frequency fluctuations characteristic of cryptocurrency markets. Higher granularity data, such as hourly or minute-by-minute prices, could provide more detailed insights and improve model accuracy. Specific model limitations were observed. For example, while Ridge Regression performed well overall, it showed potential signs of slight overfitting. This could be mitigated with further hyperparameter tuning and regularization techniques. Another notable limitation was the model’s underperformance in forecasting BTC-USD. Bitcoin’s volatility and market behavior may differ significantly from other cryptocurrencies, requiring specialized models or additional features to capture its unique patterns. The lack of high-frequency data and potential overfitting in the training phase could also contribute to these inconclusive results. This underperformance in forecasting BTC-USD suggests the need for further research and model refinement to better capture the unique characteristics of Bitcoin’s market behavior. To address these challenges, future work could focus on integrating more granular data and additional features, such as social media sentiment and market news. This would likely provide a more holistic view of the market dynamics and enhance the model’s ab ility to accurately predict price movements, instilling the audience a sense of optimism about the study’s 8.0.5 Future Work

8.0.2 Comparison with Existing Studies

While this study did not rely on any single previous study, it compared processes and results with various existing studies in the field. These comparisons revealed that the approach aligns well with standard practices and findings in the literature, instilling confidence in the audience about the research methodology. For example, previous research has supported the use of Ridge Regression for handling multicollinearity and its effectiveness in time series prediction. The results corroborate these findings, indicating Ridge Regression is particularly well-suited for cryptocurrency markets’ complex and volatile nature. This study extends existing knowledge in cryptocurrency price prediction by highlighting the importance of robust model selection and evaluation. While the results are promising, they underscore the need for more advanced methodologies to achieve greater accuracy. Incorporating sentiment analysis from social media platforms like Reddit and Twitter could capture additional volatility and hyper-bullish moments, which refer to periods of extremely positive market sentiment and high buying activity. These moments often precede significant price increases, making them essential for accurate price prediction. Moreover, combining technical indicators with sentiment data could significantly enhance the model’s predictive power and reliability. 8.0.3 Extension of Existing Knowledge

88