Predicting Cryptocurrency Exchange Rates
summary
A streaming engine ingests Twitter and exchange feeds, applies sentiment and feature windows, trains six to seven algorithms per coin online, and hot‑plugs best models into production; it runs across Kafka streams and microservices and evaluates 80 cryptocurrencies continuously. It achieves 70–90% success vs Bitcoin for some coins but fails to overcome spreads, commissions and market collapse that reduce signals, so the model works relative to Bitcoin yet does not reliably generate net profit.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:31Background
- 1:14Why Forecast Cryptocurrency
- 3:14How Crypto Ecosystem Works
- 5:43How Do We Teach Our Mind
- 6:49Concept Drifts
- 9:36Different Exchanges
- 12:17The Architecture
- 14:09The Model
- 15:09Historical Situation
- 15:37Training Streams
- 19:04Kafka Engine
- 21:02Results
- 24:47Problems
- 28:33What we learned
- 31:52Summary
related talks
40mSleeping Through the Volatility: Building an AI Agent for Automated Crypto Trading
35mCan Memes Predict Markets? Using Social Media Sentiment for Stock Forecasting
30mHow to build a stock market money printing machine with AI
48mScaling a cryptocurrency exchange using Data Science & Machine learning
45mTame Crypto Events with Scala and Akka Streams
30mPredictive trading
24mHow Bloomberg Uses AI to Help Speed up Over-the-Counter Trading
30mIncreasing the trading prediction by mining aggregated human texting messages