🎭

Feature Engineering For Quantitative Models

As I continue to build out my algorithmic trading models, here is an ongoing list of features I find would be helpful in elucidating high probability trade setups. These may or may not be correct — but these are the features I am experimenting with.

  • Analyzing tick data on a stock’s price
    • Close, open, high, low prices
    • Trade volume
    • Volume profile distribution
  • Generating real-time averages
    • Simple Moving Average (SMA)
      • Particularly a 20-day window
    • Exponential Moving Average (EMA)
      • Particularly an 8-day window
  • Gauging any momentum
    • Relative Strength Index (RSI)
      • Specifically margins for x where x ≤ 30 or x ≥ 70
    • Moving Average Convergence Divergence (MACD)
  • Assessing volatility
    • Historical volatility
    • Implied volatility
  • Baking in the news
    • Earnings announcements
    • Mergers and acquisitions
  • Analyzing order books
    • Focus on orders that were never executed
    • Isolate any large trade volume outliers in an order book
    • Determine whether the standard deviation of unexecuted order book trades exceeds a certain k-factor
    • Assess for any skewness
    • Apply kurtosis
    • Bid-ask spread (seeking any imbalances)
    • Order book depth
  • Sentiment analysis (opinion mining)
  • Correlation with other assets
    • Correlation coefficient with relevant indices or assets
  • Gauge the health of the economy
    • Employment data
      • Unemployment rates
    • GDP growth
      • Consumer spending
      • Industrial production
    • Inflation rates
    • Use short-term and long-term interest rates to generate a unique score
  • Find a temporal correlation with equity prices
    • Time of day
    • Day of the week
    • Seasonal trends
  • Incorporating and equity’s calculated risk factor
    • Value at Risk (VaR)
    • Conditional Value at Risk (CVaR)

Examples of features I’ve heard of but won’t use:

  • Analyzing satellite data showing the parking lots of shopping malls or factories to analyze activity levels and extrapolate projections around business activity
  • Analyzing the text of news reports
  • Searching for corporate action announcements