Classical Methods

Overview

The classical methods module provides implementations of traditional machine learning algorithms for tabular data analysis. All methods inherit from the base classical_methods class and provide a unified interface for training, prediction, and evaluation.

Available Methods

Classification and Regression Methods: - Support Vector Machine (SVM) - XGBoost - K-Nearest Neighbors (KNN) - Random Forest - LightGBM - CatBoost

Classification Only Methods: - Logistic Regression - Naive Bayes - Nearest Centroid Method (NCM)

Regression Only Methods: - Linear Regression

Baseline Methods: - Dummy Classifier/Regressor

Common Features

All classical methods in TALENT share the following features:

  • Automatic data preprocessing (missing value handling, encoding, normalization)

  • Unified evaluation metrics

  • Model persistence (save/load functionality)

  • Support for both numerical and categorical features

  • Configurable hyperparameters

  • Training time measurement

Usage Example

from TALENT.model.classical_methods import SvmMethod

# Initialize the model
svm = SvmMethod(args, is_regression=False)

# Train the model
time_cost = svm.fit(data, info, train=True)

# Make predictions
vres, metric_name, predictions = svm.predict(test_data, info, model_name)

This section contains documentation for all classical machine learning methods implemented in TALENT.

Contents: