GeoZ 2.0.1

Changelog for GeoZ 2.0.1

We're excited to announce the release of GeoZ 2.0! This version introduces significant updates and new features that warrant the major version increment. For a detailed review of changes, please refer to the Updates to prepare for 2.0 release commit.

Major Updates

1. New Function: voronoi_regions_plot

   • Specializes in mapping datasets with a small number of samples
   • Based on collaborative work with researchers from Hohai University (Fang et al., 2024)
   • Supports masking for specific geographical areas
   • Implements consistent color mapping with other functions
   • Includes options for point display and transparency

2. Global Settings

   • Introduced set_geoz_options() function to modify global settings for all GeoZ plotting functions
   • Configurable color scheme and figure resolution

3. Automatic Aspect Ratio Calculation

   • Implemented mechanism to display maps with correct geographical proportions
   • Enabled by default for more natural-looking visualizations

4. Expanded Default Colormap

   • Changed from "Set3" to "tab20", allowing representation of up to 20 distinct clusters
   • Note: mlx_plot uses its own colormap based on the MLextend library

5. Jupyter Notebook Examples

   • Added example notebooks for each GeoZ function, demonstrating capabilities and parameters

6. Synthetic Dataset

   • Generated using drawdata library
   • Includes a simplified map of the United Arab Emirates (UAE) created with ArcGIS Pro

7. GIS Convention Compliance

   • Now expects latitude in the first column and longitude in the second, regardless of column names

8. Performance Optimization

   • Added n_jobs parameter to mlx_plot for parallel processing
   • Implemented via PR to MLextend library: PR #998

General Updates

1. convex_hull_plot Function Enhancements

   • Improved color mapping consistency
   • Added show_points parameter
   • Implemented automatic figure size calculation
   • Removed dependency on descartes, now using matplotlib's Polygon
   • Enhanced error handling for clusters with fewer than 3 points

2. sklearn_plot Function Improvements

   • Added support for custom axes (ax parameter)
   • Introduced alpha parameter for transparency control
   • Enhanced bazel functionality with improved output

3. mlx_plot Function Updates

   • Added support for custom axes (ax parameter)
   • Improved color mapping consistency
   • Enhanced bazel functionality

4. bazel_cluster Function Refinements

   • Improved robustness and flexibility
   • Enhanced compatibility with pandas Series input for y_pred
   • Better handling of latitude and longitude column names

5. Documentation Improvements

   • Expanded function descriptions and parameter explanations
   • Added usage notes and examples
   • Improved consistency in documentation style (NumPy Style)

6. Code Structure and Style Enhancements

   • Implemented more consistent naming conventions
   • Improved error handling and user feedback
   • Enhanced code readability

7. Functionality Enhancements

   • Improved handling of different input data formats
   • Better integration with matplotlib
   • Enhanced compatibility with various y_pred input types (list, numpy array, pandas Series)

Minor Updates

1. Dependency Changes

   • Removed descartes dependency, replaced with direct use of matplotlib's Polygon

2. Pandas 2.x Compatibility

   • Updated internal code to support pandas 2.x API changes

3. Setup File Update

   • Updated to include new functions and dependencies

Breaking Changes

• Removed explicit support for 'LATITUDE' and 'LONGITUDE' column names, now using positional indexing

---

We encourage users to explore the new features, particularly the voronoi_regions_plot function. If you use this function in your research, please cite Fang et al., 2024. As always, we welcome feedback and contributions from our user community.

1.6.3

• Preparation for Publication
• Created a DOI for the repository
• Minor Bug Fixes

v1.6.2

This release is a major milestone in GeoZ as we open sourced the library code (it was available as pypi package since December). This release also includes the following:

• We changed the License of the Library to accommodate the source code release by changing the License to the popular BSD-3
• We wrote initial documentation on the github along with the steps to install and run GeoZ.
• Surfaced crs parameter which controls the Geographic Reference System for any points drawn in sklearn_plot.
• Fixed a few bugs or mistakes in the docStrings that were too minor to mention in the release notes.

v1.5.1

Fixed an Indentation issue in the DocStrings.

v1.5

• A major weaknesses of the model-based plotting functions in GeoZ is their inability to draw any spatially overlapping clusters.
• Due to the nature of the drawing algorithms, they can not detect nor plot any spatially overlapping clusters, We hope to address this issue directly in the future with a Fuzzy implementation, instead of the current hard classification approach.
• We have added "Convex Hull" plotting Function as a primitive and easy way to illustrate the clusters spatially.
• As this function was implemented for illustrative purpose, we do NOT recommend using it to produce any actual maps since its constricted to produce geometrical shapes only, which doesn't reflect any kind of natural phenomena.

1.4

• surfaced the SVC hyperparameters to allow the user to customize the model uncertainty.
• added appropriate docstrings explaining the parameters and their effect on the drawing process.
• cleaned the code a bit
• Updated the Library Description to be more accurate and easier to find

1.3.1

bug fixes
streamlined updating PyPi repository
updated the Readme.md

initial public release

• update to support all recent dependencies
• updated the API to be easier to understand
• updated the DocStrings to include bazel_cluster details
• surfaced the bazel_cluster details
• hidden the SVM parameters until 1.5 release.
• first public release
• pypi package release