All notable changes to SciWizard are documented here.
Format follows Keep a Changelog. Versioning follows Semantic Versioning.
- Data panel — CSV loading, table preview (up to 500 rows), data profiling summary, missing value handling (drop, mean, median, mode fill, reset)
- Preprocessing panel — label encoding, one-hot encoding, column dropping
- Visualization panel — histogram, scatter plot, correlation heatmap, feature distributions, PCA 2D projection with class colouring
- Training panel — 7 classification algorithms, 7 regression algorithms, configurable train/test split, StandardScaler toggle, k-fold CV scores, threaded execution
- AutoML panel — automatic sweep of all catalogue models, sortable leaderboard, best model highlight
- Evaluation panel — confusion matrix with cell annotations, ROC curve (binary and multi-class OvR), cross-validation bar chart
- Prediction panel — form-based single-row prediction, batch CSV prediction with export
- Model Registry — joblib persistence, metadata JSON, alias support, list/load/delete UI
- Experiment Tracker — JSONL-backed run log, full metrics, CV stats, dataset name, notes; clearable from UI
- Plugin system — dynamic Python module loading from
/pluginsat startup; supports custom models and preprocessors - Dark theme — Catppuccin-inspired stylesheet, custom Qt palette, embedded matplotlib dark style
- Windows integration — AppUserModelID via ctypes, taskbar icon from
icon/icon.ico - Beginner mode toggle — sidebar switch (foundation for future contextual help)
- Non-blocking UI — all training and batch operations run in
QRunnable/QThreadworkers - Full test suite covering
DataManager,ModelTrainer,ModelRegistry,ExperimentTracker
- Hyperparameter grid search UI with
GridSearchCV - SHAP feature importance panel
- Export trained model as standalone Python script
- Light theme option
- Stratified k-fold toggle
- Import/export experiment history as CSV