"Trading System - Feature Engineering - Part 1"

Here’s the full table with all the features we discussed, including custom features and TA-Lib indicators:

Feature Custom Feature/Calculation TA-Lib Indicator/Function Description

Log Returns log_return = np.log(close / close.shift(1)) NA Measures daily return as the logarithmic difference between consecutive closing prices.

Cumulative Returns cumulative_return = (1 + log_return).cumprod() - 1 NA Cumulative return over time based on daily log returns.

Rolling Volatility (Std Dev) rolling_volatility = df.groupby(‘isin’)[‘log_return’].transform(lambda x: x.rolling(window=14).std()) NA Rolling standard deviation of returns, capturing volatility over a specific window.

Rolling Momentum (Mean Return) rolling_momentum = df.groupby(‘isin’)[‘log_return’].transform(lambda x: x.rolling(window=14).mean()) NA Measures the average momentum over a rolling window of time.

14-day Rolling Max Price rolling_max = df.groupby(‘isin’)[‘close’].transform(lambda x: x.rolling(window=14).max()) NA Maximum closing price over the past 14 days.

14-day Rolling Min Price rolling_min = df.groupby(‘isin’)[‘close’].transform(lambda x: x.rolling(window=14).min()) NA Minimum closing price over the past 14 days.

Price Range (Close - Low) price_range = close - low NA Difference between the closing price and the lowest price of the day.

Volume-based Percentage volume_pct = df.groupby(‘isin’)[‘volume’].transform(lambda x: x / x.rolling(window=14).sum()) NA Measures the stock’s volume as a percentage of its total trading volume over a window.

Rolling Trading Volume rolling_total_trading_volume = df.groupby(‘isin’)[‘totaltradingvolume’].transform(lambda x: x.rolling(window=14).sum()) NA Rolling sum of the total trading volume over the past 14 days.

Rolling Trade Value rolling_total_trade_value = df.groupby(‘isin’)[‘totaltradevalue’].transform(lambda x: x.rolling(window=14).sum()) NA Rolling sum of the total trade value over the past 14 days.

Rolling Number of Trades rolling_total_number_of_trades = df.groupby(‘isin’)[‘totalnumberoftradesexecuted’].transform(lambda x: x.rolling(window=14).sum()) NA Rolling sum of the number of trades executed over the past 14 days.

Simple Moving Average (SMA) NA SMA(close, timeperiod=7) A moving average of the closing price over a fixed period (e.g., 7 days).

Exponential Moving Average (EMA) NA EMA(close, timeperiod=14) A weighted moving average giving more weight to recent prices.

Relative Strength Index (RSI) NA RSI(close, timeperiod=14) Measures the magnitude of recent price changes to evaluate overbought/oversold conditions.

Average True Range (ATR) NA ATR(high, low, close, timeperiod=14) Measures volatility by decomposing the entire range of price movement.

Bollinger Bands NA BBANDS(close, timeperiod=20, nbdevup=2, nbdevdn=2, matype=0) A band plotted above and below a moving average, capturing volatility.

Moving Average Convergence Divergence (MACD) NA MACD(close, fastperiod=12, slowperiod=26, signalperiod=9) Shows the relationship between two moving averages of a stock’s price.

Stochastic Oscillator (Stoch) NA STOCH(high, low, close, fastk_period=14, slowk_period=3, slowd_period=3) Indicates overbought/oversold conditions by comparing a stock’s closing price to its price range over a period.

On-Balance Volume (OBV) obv = (df[‘volume’] * np.sign(df[‘close’].diff())).cumsum() OBV(close, volume) A cumulative volume-based indicator that shows whether volume is flowing into or out of a stock.

Moving Average of Volume (MAV) mav = df[‘volume’].rolling(window=14).mean() NA Moving average of volume over a rolling window.

Relative Volume (RVOL) rvol = df[‘volume’] / df[‘volume’].rolling(window=14).mean() NA Measures volume relative to the average volume over a set period.

Price-to-Earnings Ratio (P/E) pe_ratio = df[‘close’] / df[‘earnings’] (if earnings data is available) NA Ratio of the stock’s price to its earnings per share.

Price-to-Book Ratio (P/B) pb_ratio = df[‘close’] / df[‘book_value’] (if book value data is available) NA Ratio of the stock’s price to its book value.

Moving Average of Volume-to-Price Ratio mav_vpr = df[‘volume’] / df[‘close’].rolling(window=14).mean() NA Measures the relationship between volume and price, averaged over a rolling window.

14-day Average of Price Range price_range_avg = df.groupby(‘isin’)[‘price_range’].transform(lambda x: x.rolling(window=14).mean()) NA The average range between closing price and low price over a rolling window.

Data Engineering of Volume

Techical Indicators.

1. Volume profile.
   1. key level or key zone
      1. Can we converted into a feature
2. Stochastic Indicator
   1. %D
   2. %K
      1. Cross over method
3. Add candle information as well
4. Don Chain Channels
   1. Best for support and resistance
5. Anchored VWAP
6. Bollinger bands
7. macd line
8. ichimoku cloud
9. fibonacci
10. RSI Diver

Lagged Features

Need to check how many lags do we need for 30 dats timeframe. Use ACP or PACF for testing the significance.

Decomposition

1. trend with lags
2. seasonalty woth lags
3. noise