The increasing popularity of bitcoin motivated us to create a platform which would facilitate trading the cryptocurrency.
What it does
By providing an algorithm which predicts short term bitcoin fluctuations, it is now easier for people without experience in such domain to start trading. They will find our project to be a reliable assistant when it comes to buying/selling as well as planning future business moves, if the context allows.
How we built it
While part of the team was busy training a Neural network designed to offer insights based on historic data and news analysis, the other was focusing on sending (and rendering) the entire analysis result spit from the NN. The server is represented by a rather basic DigitalOcean droplet that was running a python container focused on both performance (zero-downtime during redeployment) as well as running async jobs to keep the platform updated with the latest analysis results that change due to our continuing working news crawler.
Challenges we ran into
The data mining task was long and painful, but we think the results are worthwile. Besides this, deployment was more challenging than expected as we had multiple scripts that had to run in parallel.
Accomplishments that we're proud of
We are rather proud of the whole project, having managed to gather data and carry out 2 Machine Learning mini-projects (data for news sentiment analysis and data related to the bitcoin blockchain) and, furthermore, the fact that we managed to deploy a reliable stack to support all the requirements that we have set on.
What we learned
Pretty much improved on all areas which the project involved.
What's next for BitcoinNewsTrader
Creating our own sentiment analysis tool rather than using a third party, finishing the online deployment and coming up with a mid-term prediction algorithm too.