First Release

[0.0.1] - 2025-11-21

This release delivers the first complete implementation of the Interval Type-2 TSK (IT2-TSK) intelligent prediction system, including automated fuzzy model construction, parameter tuning, and evaluation.

Added

• Full implementation of the Interval Type-2 TSK (IT2-TSK) prediction model and IT2-Mamdani baseline.
• Automated generation of membership functions (MFs) and rule bases using Mini-Batch K-Means and Gaussian MFs.
• Two-stage parameter tuning:
• Grid Search (GS) for coarse exploration of parameter ranges (UF, MSR, min-std ratio, etc.).
• Genetic Algorithm (GA) for adaptive fine-tuning of MF parameters.
• Complete preprocessing pipeline: normalization, feature selection, and stratified train/validation/test splitting.
• Automatic evaluation module: MSE, RMSE, MAE, R², and prediction vs. ground-truth visualizations.
• Modular project structure: src/, scripts/, configs/, tests/, notebooks/.
• Documentation of experimental settings and model parameters.

Changed

• Consolidated fuzzy logic modules, clustering, rule generation, and parameter tuning into a unified pipeline.
• Updated experiment configuration format to align with IT2 parameterization (UF, MSR, etc.).

Known Issues

• Genetic Algorithm tuning is stochastic, so results may vary between runs.
• Triangular MFs showed less stable performance compared to Gaussian MFs on several datasets.
• Excessively high UF values may produce overly wide FOUs, reducing prediction accuracy.

Notes

• Average model performance across 12 datasets: R² ≈ 0.98.
• Full methodological details and parameter tables are available in the thesis document and paper.