NEWS

---

deaviz 0.1.0 (2026-07-02)

Initial release. deaviz provides high-dimensional visualization methods for Data Envelopment Analysis (DEA), built around a single validated dea_data() object and following a compute_*() / plot_*() naming convention.

Data object

• dea_data() constructs the validated input/output object that every function consumes; as_dea_data() coerces existing data. print() methods are provided for the dea_data and dea_som classes.

Analysis

• compute_efficiency() --- radial DEA efficiency scores (CRS, VRS, DRS, IRS or FDH; input- or output-oriented).
• compute_cross_efficiency(), compute_cross_efficiency_weights() and standardize_weights() --- cross-efficiency scores and their weight profiles.
• compute_multiplier_weights() --- optimal input/output multipliers.
• compute_som() --- a self-organizing map of the input/output profiles.

Visualization

• Data and efficiency overview: plot_io_distributions(), plot_efficiency_distributions(), plot_io_efficients(), plot_io_scatter(), plot_io_heatmap().
• Frontier and projections: plot_io_costa_frontier(), plot_io_pca_biplot(), plot_io_mds(), plot_io_3dscatter().
• Benchmarking networks: plot_io_lambda_network(), plot_io_peer_network().
• Cross-efficiency: plot_cem_heatmap(), plot_cem_unfolding(), plot_cem_weights_heatmap().
• Profiles: plot_io_radar() (with its coord_radar() coordinate system) and plot_io_parcoo().
• Self-organizing maps: plot_io_som(), plot_io_som_components().
• Panel data: plot_panel_io_biplot() draws each DMU's trajectory over time.

Cross-cutting features

• Passing a single DMU to labels fades the rest of the plot into a focus view; the fade argument tunes the level or disables it.
• x_angle rotates long x-axis tick labels on plot_io_distributions(), plot_io_heatmap(), plot_io_parcoo(), plot_cem_heatmap() and plot_cem_weights_heatmap().
• Many plots accept interactive = TRUE to return a plotly widget.
• Consistent, colour-blind-safe style throughout: the Okabe-Ito qualitative palette, a viridis sequential palette, and a shared minimal theme.

Datasets

• chinese_cities --- 35 Chinese cities with three inputs and three outputs (Sueyoshi, 1992).
• taiwanese_banks --- a balanced panel of 22 Taiwanese commercial banks over 2009-2011 (Kao & Liu, 2014), the worked example for plot_panel_io_biplot().