rBitrage

Using machine learning and Coinbase's API, we are analyzing and predicting the relationship between BTC and ETH

Comment

Inspiration

We are a group of firm believers that think cryptocurrency is going to have a huge impact in both business and everyday life — this project allowed us to get hands-on experience in understanding what makes them tick.

What it does

Our model predicts how movement in BTC will influence the price movement of ETH in the future, especially in a minute, five minutes, and thirty-minute interval. While also, allowing certain events such as volume inflow or outflow to predict price movement as well.

How we built it

We gathered together historical data of prices for ETH and BTC every minute over the past 6 months using Coinbase's API. Then, we used python to accumulate and smooth the data for gaps and missing information. Then we used Microsoft Azure ML Studio to analyze various features, for example, the change in volume over time of each cryptocurrency relative to each other. Once a predictive model was built by the ML Studio, we backtested September data that was previously not used in the training of the model. Using C++ to run the model as if we were actively trading in real time. Then an iOS app was created that allows users to view current prices of ETH and BTC along with what our model predicts the price movement in set future intervals.

Challenges we ran into

When coming up with features for the machine learning platform, we ran into issues specifically in how we wanted our data to be analyzed. For example, our initial testing placed the same weight on historical data from 6 months ago as recent data from a couple days ago. We learned that this did not optimize our solution and went back to the drawing board. With data relevance in mind, we used key financial statistics typically used for stock and portfolio analysis. As well as coming up with some statistics in search of a pattern between BTC and ETH.

Accomplishments that we're proud of

This project underwent a huge change after the first day. We originally had the idea of creating an iOS application that allowed people to invest their spare change from transactions directly into cryptocurrencies. After working on our app, we realized that we wanted to focus more on the statistical analysis of the market for cryptocurrencies.

What we learned

None of our group members had experience with linking API's with iOS before, so a lot of time was spent debugging and making sure everything was running smoothly. Most of our experience has been working within the bounds of a school project and finally, we were able to learn just how difficult it is to accumulate, manipulate, and analyze a real-life data set.

What's next for rBitrage

This weekend has been an incredible learning experience, but we hope to continue this project by fully implementing a real trading platform based on revised models soon. We also look forward to learning more about the cryptocurrency space and how it can change the future.

Share this project:

Updates