v1.2.7

Release v1.2.7

Overview

This release brings significant improvements to OpenDSM's clustering methodology, modernizes the build system, and integrates comparison group functionality although it is not currently documented or exposed cleanly through the API.

New Features

Enhanced Clustering with Multi-Metric Voting System

• Clustering Revision - Introduced a new voting system that evaluates multiple clustering indices to determine optimal cluster configurations (#580)
  • Supports voting between various clustering quality metrics
  • Improves cluster selection reliability and robustness
  • Adds comprehensive tests for the new clustering methodology

Comparison Groups

• Comparison Group Functionality - Integrated Comparison Groups into OpenDSM, providing basic access to comparison groups (#577)
  • Consolidates clustering methodology across systems
  • Currently undocumented code and we still need to develop API for user ease-of-use
  • Includes 176 commits of GRIDmeter codebase integration

Improvements

Build System Modernization

• Migration to pyproject.toml - Replaced deprecated setup.py with modern pyproject.toml configuration (#566)
  • Switched to uv package manager for improved dependency management
  • Addresses Docker build deprecation warnings (setup.py support ending Oct 2025)
  • Updated Dockerfile to use uv pip compile and uv pip install

Bug Fixes

• Example Data Fix - Corrected temperature values in example data that were accidentally modified (#575)
• Clustering Bug Fixes - Various refinements and bug fixes in clustering implementation
• Dependency Updates - Fixed and updated project dependencies

Testing

• Added new tests for clustering voting system
• Enhanced test coverage for clustering methodology
• Updated test platforms and configurations

Full Changelog: v1.2.6...v1.2.7

v1.2.6

What's Changed

• Fixed bug in from_series instantiation of daily data class by @travis-recurve in #565

Full Changelog: v1.2.5...v1.2.6

v1.2.5

What's Changed

• Expose SpectralClustering's assign_labels options. discretize and cluster_qr are not always deterministic with seed so for regression testing kmeans is suggested by @travis-recurve in #564
• Added more metrics to BaselineMetrics by @travis-recurve in #564

Full Changelog: v1.2.4...v1.2.5

v1.2.4

What's Changed

• Added seeding to spectral clustering by @travis-recurve in #560
• Fixed bug where seed was not being propagated to clustering settings by @travis-recurve in #560
• New model metrics by @travis-recurve in #561
• Changed failed model metrics to be nan rather than None by @travis-recurve in #561

Full Changelog: v1.2.2...v1.2.3

v1.2.2

What's Changed

• Fixed daily model not using adjusted normalized metrics by @travis-recurve in #559
• Ensure the autocorrelation function usage is consistent by @travis-recurve in #559

Full Changelog: v1.2.1...v1.2.2

v1.2.1

What's Changed

• Revert how autocorr is calculated by @canchola-recurve in #558

New Contributors

• @canchola-recurve made their first contribution in #558

Full Changelog: v1.2.0...v1.2.1

v1.2.0

What's Changed

• Add hourly model uncertainty by @travis-recurve in #553
• Daily and billing models now use BaselineMetrics natively by @travis-recurve in #554
• Data sufficiency can now be modified using a settings dictionary for R&D purposes by @travis-recurve in #555
• Migrate to modern logger interface by @emmanuel-ferdman in #551

New Contributors

• @emmanuel-ferdman made their first contribution in #551

Full Changelog: v1.1.0...v1.2.0

v1.1.0

What's Changed

Hourly Model

• disallow negative cvrmse in hourly model
• added temp_bin/cluster/temp interaction
• updating clustering
• updating hourly settings and adding clustering
• clustering sort update
• adaptive robust elasticnet
• added fixed temperature bins
• extreme temperature bin settings update
• modified temporal_cluster code

Daily Model

• added daily CVRMSE >= 0 and PNRMSE pass for sufficiency
• fixing broken tests and making daily model use baseline_metrics
• updating daily model settings to no longer use model_fields on instantiated class

Miscellaneous

• revised adaptive loss penalization
• modified extreme value flag to be based on IQR rule
• ignores numpy warnings when dividing by zero and returns a NaN instead by @ssuffian in #549
• fixes the warning data to only inculde invalid temperature rows by @ssuffian in #548
• update README.md by @travis-recurve in #546

Full Changelog: v1.0.0...v1.1.0

v1.0.0

Initial release of OpenDSM