params.yaml

# Configuration file for DVC (Data Version Control)

# This YAML file defines the parameters used in the DVC pipeline for training a machine learning model.
# It includes data paths, training configurations, and model settings.

# Data paths for training and testing
data:
  # Path to the training data in Parquet format
  train: data/train.parquet

  # Path to the testing data in Parquet format
  test: data/test.parquet 

# Training configuration
train:
  # Number of hyperparameter optimization trials
  n_trials: 40
  
  # Hyperparameter ranges for Optuna
  hyperparams:
    # Maximim n-grams for the first TF-IDF component (word)
    max_ngram_word:
      low: 1
      high: 3

    # Maximim n-grams for the second TF-IDF component (char)
    max_ngram_char:
      low: 1
      high: 5

    # Loweracase option for TF-IDF
    lowercase:
      - True
      - False

    # Use IDF
    use_idf:
      - True
      - False

    # Regularization parameter C for Linear SVC
    C:
      low: 1e-7
      high: 10

    # Tolerance for stopping criteria
    tol:
      low: 1e-5
      high: 1e-1

    # Loss function
    loss:
      - hinge          
      - squared_hinge

    # Method used for calibration
    calibration:
      - isotonic
      - sigmoid

# Model paths
model:
  # Path to save the model in pickle format
  pickle: live/model/model.pkl 

  # Path to save the model in ONNX format
  onnx: live/model/model.onnx