Skip to content

CHANGELOG.md

Latest commit

 

History

History
143 lines (99 loc) · 4.31 KB

CHANGELOG.md

File metadata and controls

143 lines (99 loc) · 4.31 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.3.0] - 2025-12-14

Added

  • Gradient Boosting Algorithms

    • LightGBM trainer with classification/regression support
    • CatBoost trainer with automatic categorical feature handling
    • Early stopping, feature importance, and native model exports

  • Clustering Algorithms

    • K-Means trainer with silhouette, calinski-harabasz, davies-bouldin metrics
    • DBSCAN trainer with automatic noise detection
    • Optimal cluster number finder (elbow method) for K-Means
    • Optimal eps finder for DBSCAN

  • Anomaly Detection Algorithms

    • Isolation Forest trainer for efficient outlier detection
    • One-Class SVM trainer with RBF kernel support
    • Anomaly scoring and threshold tuning capabilities
    • Support for supervised evaluation with ground truth labels

  • Example Configurations

    • Added lightgbm.json and catboost.json configs
    • Added clustering/kmeans.json and clustering/dbscan.json
    • Added anomaly/isolation_forest.json and anomaly/one_class_svm.json

Changed

  • Updated trainer registry with 6 new algorithm types
  • Extended model type support: classification, regression, clustering, anomaly_detection

[0.2.0] - 2025-12-07

Added

  • Documentation

    • Created comprehensive docs/ folder with getting-started guide, installation instructions
    • Added examples/ folder with sample configs for all model types (RF, XGB, Logistic, DNN)
    • Added example tuning configurations for hyperparameter optimization
    • Added CONTRIBUTING.md with detailed contribution guidelines
    • Added CODE_OF_CONDUCT.md for community standards
    • Added SECURITY.md for vulnerability reporting

  • Testing Infrastructure

    • Created tests/ folder with pytest framework
    • Added test suite for trainers (RF, XGBoost, Logistic Regression)
    • Added test suite for tuners (Grid Search, Random Search)
    • Added shared fixtures and configuration in conftest.py

Changed

  • Cleaned up project structure for production deployment
  • Updated .gitignore to exclude generated artifacts properly
  • Improved GridSearchTuner to handle n_trials parameter gracefully

Removed

  • Hotel-specific datasets and configurations (moved to generic examples)
  • Test experiment artifacts from repository
  • Empty scripts/ directory

Fixed

  • GridSearchTuner now properly filters out unsupported n_trials parameter
  • Configuration files now use generic data/your_data.csv paths

[0.1.0] - 2025-12-03

Added

  • CLI Training Pipeline

    • Train ML models (Logistic Regression, SVM, Random Forest, XGBoost)
    • Train DL models (TensorFlow DNN, CNN, RNN/LSTM/GRU)
    • Configuration-driven training via JSON/YAML files
    • Parameter overrides from command line

  • Hyperparameter Tuning

    • Grid Search for exhaustive parameter search
    • Random Search for large parameter spaces
    • Bayesian Optimization via Optuna for intelligent search
    • Cross-validation support
    • Auto-train best model after tuning

  • Model Explainability

    • SHAP (SHapley Additive exPlanations) for global/local explanations
    • LIME (Local Interpretable Model-agnostic Explanations)
    • Feature importance visualization
    • Instance-level explanations

  • Data Preprocessing Pipeline

    • Scaling: StandardScaler, MinMaxScaler, RobustScaler
    • Normalization: L1, L2, Max norm
    • Encoding: LabelEncoder, OneHotEncoder, OrdinalEncoder
    • Feature Selection: SelectKBest, RFE, VarianceThreshold
    • Pipeline support for chaining preprocessors

  • Experiment Tracking

    • Automatic experiment logging
    • Run comparison and filtering
    • Export to CSV
    • Mini-MLflow style tracking

  • Model Export

    • ML models: Pickle, Joblib, ONNX
    • DL models: SavedModel, H5

  • Interactive Terminal UI (TUI)

    • Train models with guided interface
    • Evaluate saved models
    • Browse experiment history
    • View registered models

  • Model Registry

    • Automatic model discovery
    • Pluggable trainer architecture
    • Easy extension for custom models

Technical Details

  • Python 3.8+ support
  • Type hints throughout codebase
  • Rich CLI output with colors and tables
  • Comprehensive error handling