Bitsight

Title Page
Process Overview
Tech Stack
Historical Data
1 Year Data
1 Month Data

Strategic Trading Track

Inspiration

The inspiration for Bitsight stemmed from observing the volatile and unpredictable nature of the cryptocurrency market, particularly Bitcoin. As Bitcoin continues to gain mainstream acceptance among government agencies and major financial institutions, it became clear that there was a need for a tool to help investors navigate its rapid price fluctuations. Our goal was to create software that leverages machine learning to provide accurate trend predictions, empowering both novice and experienced investors—whether individuals or corporations—to make more informed decisions in this evolving financial landscape.

What it does

Bitsight serves as a Bitcoin trend predictor that analyzes historical data to forecast future price movements. By employing machine learning algorithms, it provides insights that help investors make strategic decisions in an increasingly volatile market.

How we built it

Building Bitsight required a detailed process of data collection and analysis. First, we gathered historical Bitcoin data from Yahoo Finance, including daily prices (open, high, low, and close) and trading volume since 2014. This gave us the foundation for understanding past trends.

To train our predictive model, we tested its accuracy by removing one week of data and predicting up to the current date. We then compared the predicted outcomes with actual price movements and adjusted the mix of data intervals to improve accuracy. This process of fine-tuning helped us build a reliable model for forecasting future trends.

We used Python and it’s libraries for this project.

Challenges we ran into

Time was our biggest challenge. Training our model took a significant amount of time — our first run lasted 50 minutes. Although the time decreased with each run, it was still a lengthy process. In hindsight, we should have settled for a lower accuracy rate to save time to create a front-end.

Another challenge we faced was determining which data and time intervals—yearly, monthly, weekly, daily, or hourly—would provide the most accurate predictions. We experimented with different combinations to find the one that best matched current Bitcoin trends, but this process required a lot of trial and error.

We initially considered incorporating sentiment analysis into Bitsight to provide deeper insights into the factors driving Bitcoin’s price movements. However, due to time constraints and the complexity of implementing such a feature, we decided not to include it in this version of the project. Instead, we focused on refining the core prediction model. Sentiment analysis requires extensive data collection and processing, which would have stretched our timeline significantly. In future iterations, we may explore adding sentiment analysis to enhance the platform’s predictive power.

Accomplishments that we're proud of

We are proud to have gotten our model’s accuracy to 97% with only a $500 difference between the actual data and predicted data.

What we learned

Machine learning is tough!

What's next for Bitsight

Looking ahead, we have several goals for improving and expanding Bitsight. We plan to enhance the accuracy of our predictions by continuing to train our model with larger datasets and more sophisticated algorithms. As we gain a deeper understanding of which data intervals yield the best results, we can fine-tune the model for greater precision, ensuring our predictions are as reliable as possible.

In addition to improving prediction accuracy, we want develop our own sentiment analysis model. While we do not currently incorporate sentiment analysis, we recognize its potential to provide valuable insights into the factors driving Bitcoin’s price movements. Our custom model will be designed to cover the entire history of Bitcoin, allowing us to analyze historical events and their impact on price trends comprehensively. By understanding these influences, we can offer users a richer context for their investment decisions.

We also plan to add a front end with a user interface, making it cleaner, more intuitive, and easier to navigate for all types of investors. A key feature we want to add is the ability for users to trade directly through the Bitsight platform. This will allow them to act on the insights they receive in real time, further integrating predictive analysis with actionable trading options.