Pre processing library - MegaJoctan/MALE5 GitHub Wiki

Preprocessing.mqh: MQL5 Normalization Techniques

This MQL5 library, preprocessing.mqh, provides three essential normalization techniques for machine learning:

  • RobustScaler
  • MinMaxScaler
  • StandardizationScaler
  • CLabelEncoder (for label encoding)

These functions help prepare your data for machine learning algorithms by scaling or encoding features to a common range or representation. This step is crucial for improving the performance and stability of many machine learning models.

Key Functionalities:

Scalers:

  • RobustScaler:
    • Uses the median and interquartile range (IQR) for scaling.
    • Less prone to outliers compared to other scalers.
  • MinMaxScaler:
    • Scales features to a range between 0 and 1.
    • Useful for algorithms sensitive to feature scales.
  • StandardizationScaler:
    • Standardizes features by subtracting the mean and dividing by the standard deviation.
    • Leads to features with zero mean and unit variance.

Label Encoder:

  • CLabelEncoder:
    • Encodes categorical string labels into numerical labels.
    • Useful for converting categorical features into numerical representations usable by machine learning models.

Usage:

Each scaler class provides:

  • fit_transform(const matrix &X): Fits the scaler on the data matrix X and performs the transformation.
  • transform(const matrix &X): Transforms the data matrix X using the fitted scaler.
  • transform(const vector &X): Transforms the data vector X using the fitted scaler.

The CLabelEncoder class offers:

  • encode(string &Arr[]): Encodes a string array into a numerical vector.

Saving and Loading Models (Scalers):

  • Each scaler class supports saving and loading models:
    • save(string save_dir): Saves the scaler model to a given directory/folder.
    • load(string dir): Loads a previously saved scaler model from a given folder.

Additional Notes:

  • Remember to include this library file (#include "preprocessing.mqh") in your MQL5 code where you want to use its functionalities.