Summary
Australian Stock Market Prediction Within 5 seconds. Stock market tick data storage and stock price prediction.
Keywords
AWS, ETL Pipelines, Tensorflow, Elasticsearch, NoSQL, Apache Kafka, Apache Flink, Python, Message Queues, Microservices (Lambda), Docker, Dask
Description
The goal was to process livestreams of tick stock market data in milliseconds and give decisions regarding trading strategies. In the further course, an automated trading bot was developed, which makes buying decisions based on machine learning.
In detail, ~120 tick data/seconds were processed, which were sent to Apache Kafka (meanwhile Flink) and then processed by autoscaling microservices (AWS Lambda). The model itself was hosted using OpenFAAS in a Docker Swarm cluster and performed purchase decisions via custom APIs to ASX.