DEV Community

Cover image for CoinSavvy: Revolutionizing Crypto Price Predictions
Vikranth U
Vikranth U

Posted on

CoinSavvy: Revolutionizing Crypto Price Predictions

Hey Coders,

Exciting times await as we guide you through CoinSavvy, a dedicated project aimed at refining Bitcoin price predictions for greater accuracy. We've encountered multiple implementation challenges, and we're eager to share our experiences and unique insights into forecasting the future value of Bitcoin.

Initial Model:

In our first attempt, we built a model centred around the first day's open value of Bitcoin. Our approach sought to measure the predictive power of this single feature in forecasting future prices. We aimed to predict the high value solely based on the open value, followed by the low value based on open and high value. Additionally, we ventured into predicting the volume value based on open, high, and low values. Finally, we forecasted the close for the day, which would become the open value for the next day.

Unfortunately, the predictions from this model weren't accurate, deviating significantly from the testing dataset.

Initial Model

Enhanced Model:

Undeterred, we refined our approach by incorporating two features: the open price and the volume traded on the respective day. This enhanced model aimed to capture additional market dynamics, resulting in predicted values that closely aligned with the actual dataset and exhibited significantly reduced deviation.

Enhanced Model

Dataset Transition:

We initially used Cryptocompare for our dataset, but we found that the data from yfinance and ccxt provided more reliability. Consequently, we made the shift, improving our predictions' accuracy.

Excitingly, we expanded our project to include USDT and Ethereum as well. This extension opens up new possibilities for accurate predictions in the ever-evolving cryptocurrency landscape.

Overall, the journey has been rewarding, and we're committed to returning, refining the prediction model, and incorporating more features.

Check out our project on GitHub. You can get hands-on experience with both models by cloning our repository. Experiment, tweak, and contribute to the evolution of CoinSavvy.

Contributors:

@vikranth3140
@fakepickle
@noeltiju
@mehul-ag20

Thank you for joining us on this exciting ride. Explore CoinSavvy today and be a crucial part of the future of crypto forecasting! 🚀📈

Top comments (0)