Price Returns Prediction Using RNN With LSTM Units

Suit Khanna, Quant Researcher, Algorithmic Trading, Edelweiss Financial Services

financial market

• Data Processing and Feature Creation

  • Input features created based on Price returns of Nifty Futures

    • inspired by Taylor expansion
    • derivatives of the returns are used as input features
    • (future: Order Book/Index Options/broad market features)

  • Output: 10 classes

• Model types and Specification

  • A Simple RNN(MLP)
  • LSTM RNN
  • Stateful LSTM RNN
  • Bi-directional LSTM RNN

• Strategy Model

  • The output of LSTM RNN can be used as input for a trading strategy model