Releases: ZJUEarthData/Geochemistrypi
Releases · ZJUEarthData/Geochemistrypi
Release list
GeochemistryPi mcp-v0.2.2-cli-v0.8.2
Signed GeochemistryPi MCP bundle. Download and extract the ZIP, then ask a supported Agent to install it. The bundled CLI is the exact wheel published to PyPI by this workflow.
GeochemistryPi mcp-v0.2.1-cli-v0.8.1
GeochemistryPi MCP 0.2.1 for CLI 0.8.1
v0.8.0
Geochemistry π v0.8.0
✨ New Features
- Multi-output Regression: Predict multiple continuous target variables simultaneously (multiple X → multiple Y) with FLAML AutoML integration
- Multi-output Classification (Experimental): Predict multiple class labels simultaneously with XGBoost native multi-strategy support
- OPTICS Clustering: Density-based clustering without pre-specifying the
epsparameter - Automated Clustering Evaluation: One-click Silhouette score calculation and visualization for K=2 to K=10
- Time Series Analysis: Independent module with customizable bin width, bootstrap iterations, and uncertainty quantification
- New Built-in Dataset:
Data_Time_Series.xlsxfor time series analysis demonstration
🔧 Improvements
- Feature selection automatically skipped for multi-output tasks
- MLflow parameter logging optimized for multi-output hyperparameters
- Visualization functions now support multi-output (each target gets separate plots)
- README enhanced with badges, structured tables, and Star History chart
🐛 Bug Fixes
- Fixed
fit()call inMultiOutputRegressorandMultiOutputClassifier - Fixed MLflow parameter length limit (500 chars) by splitting long hyperparameters
- Fixed LightGBM compatibility issues in multi-output tasks
- Fixed Python 3.9 type annotation compatibility (
dict | None→Optional[dict]) - Fixed
silhouette_scoreimport in clustering module - Resolved application data null value problem (PR #426)
- Fixed SVM and Decision Tree AutoML runtime issues (PR #429)
Geochemistrypi_v0.7.0.exe
Merge pull request #424 from ZJUEarthData/web docs: update the changelog with the info about v0.7.0.
