CoinPredictor is an advanced forecast and analysis service, offering a one-stop solution for traders and researchers. It incorporates a unique combination of signals not available together in one product from any other single source.
Incorporate bitcoin market stats like hashrate, active workers and block length.
Time Series Model predicting the next 1 min and 10 min prices.
10 minute projection based on Google Trends topic data.
Real Time flash event display.
Analyze the frequency of flash events and provide a clear signal of market instability.
There are several ways to improve on what we've built:
The public APIs we use allow three queries per minute. In order to truly be able to use our algorithms, the interval range would need to be in the miliseconds, so our ranges would need to be much smaller. To get this information, we'd need to work on a paid tier API.
We can improve our weights by implementing machine learning models that predict flash events and price movements. In this project, we limited our scope to identifying initial signals and showing a proof of concept. Now that we're able to predict a price, the next step would be to add a feedback loop so that our model would improve over time.
In order to deliver a working product, we needed to limit project scope. However there is a short-list of additional signals that could improve our model: