Web app that aims to predicts the daily and weekly top 10 cryptocurrencies using a RNN with an LSTM architecture
- Model: To gather the data, I used an API to collect historical data of a few hundreds cryptocurrencies's daily price, volume, social media comments, likes, etc. Next, I cleaned the data and regularized it with Python to account for the variations accross the data. Finally, I trained a RNN with an LSTM architecture using Keras.
- Frontend: I first used Figma to design the main page, and then used React to build it.
- Backend: I deployed the model to S3 and implemented a function in Lambda that allows me to make an HTTP request with data to the model and receive a prediciton. I used MySQL to store the daily data generated by the model. I made a script on an EC2 instance that automatically updates the predictions daily.