/combining_alpha_signals

Combines alpha signals using a random forest, and demonstrates solutions to overlapping labels issue (rolling autocorrelation) proposed by Marcos Lopez de Prado in Advances in Financial Machine Learning

Primary LanguageJupyter Notebook

Combining Alpha Signals

This project builds off the factor model project where a simple mean was used to combine the alpha signals, here:

  • additional features are added such as date, regime, stock volatilitity
  • the signals are combined using a random forest for enhanced alpha
  • demonstrates how to address overlapping labels issue (rolling autocorrelation) with DTs/Random Forests

drawing

note: end of day data from Quotemedia, and sector data from Sharadar was used which could not be shared due to licensing restrictions