For all current online services for cryptocurrency transaction, they lack some useful data for users to predict and to understand the true mechanism behind the cryptocurrency. Therefore, we added and analyzed three crucial statistics in our services. One is recent activities on github for that cryptocurrency, one is discussion in community such as comments or posts on reddit, and the other one is recent news related to that cryptocurrency. And we use those data to train the model in order to generate future prediction for the price using Azure services and CNN by Keras library.
What it does
Predict for future price of cryptocurrency. Provide data for users to analyze. Provide a discussion platform for users to discuss about cryptocurrency and put it back to ML to retrain the data.
How we built it
Firstly, we leveraged python crawler to crawl all related data in reddit, news and github. Then we fed those data to our training CNN and receive the modal. After that, we utilized the modal to predict future price based on those three major topics. Finally, combing ReactJS, Microsoft Bot Services, and Firebase to store the data and generate the frontend of our services.
Challenges we ran into
In the beginning, it's really hard for us to retrieve the data from reddit, news and github due to their format. Moreover, it really hard for us to use those unformatted data to train the model.
Accomplishments that we're proud of
Building one Machine Learning Model, one Natural Language Processing AI Chatbot and one full-stack web in 24 hrs. And they all work together!!!
What's next for Bidcoin
Expand the types of cryptocurrency. Add payment services for users to pay cryptocurrency in real time.