-
First attempt at predicting Ethereum data using a Bi-LSTM model
-
Infrastructure and Data pipeline
Inspiration
Learning about ETL pipelines, MLOps, and Trading bots in a production environment
What it does
Streams time-series data, trains a Recurrent Neural Network on initial data and future batches of data, and periodically warehouses prediction and historical data into a NoSQL database. The database is then used by a high-performant trading bot that makes buy and sell decisions.
How we built it
Python, Kafka, Spark, Keras, MongoDB, Docker
Challenges we ran into
Locally caching model data based on best validation results
Accomplishments that we're proud of
Predicted the general trends in an Ethereum time-series dataset
What we learned
Money doesn't grow on trees, but it does grow using decision trees :)
Log in or sign up for Devpost to join the conversation.