Predictive trading
summary
Product uses 105 stock and 145 bond indices across ~18 years, engineers technical and macro features (returns, volatility, Sharpe), and predicts weekly for stocks / monthly for bonds to rank assets. Models are random forest & extra trees trained both on full data and with a learning on extreme cases approach (training on top/bottom 20%), then backtested 2012–2018: prediction-based portfolio yields ~200% cumulative return vs ~70% for the equally weighted benchmark. Deployed as a daily-updating web app and used by a Slovenian insurance company.
AI-written summary · about
chapters (13)
- 0:00Start
- 0:29Predictive Trading
- 4:45Asset Universe
- 6:22Data Preparation
- 8:30Technical Features
- 11:45Target Variable
- 14:00Back Test Strategy
- 14:46Back Test Strategy Performance Metrics
- 17:14Example Model
- 20:13Live Demo
- 22:00Technology
- 22:26Limitations
- 24:30Assure that the Model Does Not Over Fit the Data
related talks
42mMachine Learning with Elixir and Phoenix
01h 26mImplementing a highly scalable Stock prediction system with R, GemFire and Spring XD
30mAI vs Physics in Portfolio Optimisation
30mIncreasing the trading prediction by mining aggregated human texting messages
21mPredicting Bond Prices: Regime Changes
56mPredictive Analytics for Practitioners
47mQuantitative Trading with Machine Learning
33mQuantamental Investing - the future of finance