mcda-viz 0.2.0

Interactive visualization dashboards for the MCDA Python ecosystem

mcda-viz

Visualization library for the MCDA Python ecosystem. Provides interactive Dash dashboards and Plotly plots for exploring and presenting MCDA results.

Built on mcda-core, maintained by the SMG research group at ULB.

Installation

pip install mcda-viz

Method-specific dashboards require the corresponding library as an extra:

pip install mcda-viz[promethee]   # PrometheeDashboard — adds mcda-promethee

Future extras will follow the same pattern as new method libraries are added to the ecosystem.

Python ≥ 3.9.

Quick start

PROMETHEE dashboard

from mcda_core import PerformanceTable
from promethee import PrometheeII
from promethee.preferences.parameters import ParametersPromethee
from mcda_viz import PrometheeDashboard

table = PerformanceTable(...)
params = ParametersPromethee(...)
result = PrometheeII(table, params)
PrometheeDashboard(result, title="My analysis").run(debug=True)

MAUT dashboard

from mcda_methods import MAUT
from mcda_viz import MAUTAnalysisDashboard

maut = MAUT(table, params)
MAUTAnalysisDashboard(maut, title="MAUT Analysis").run(debug=True)

See example_dashboards/ for complete runnable examples with real datasets.

What's included

Dashboards

class	description
PrometheeDashboard	GAIA plane + net flows parallel coordinates plot; live recomputation when weights or preference functions change
MAUTAnalysisDashboard	Parallel coordinates in evaluation and utility space; criterion-wise A vs B comparison; utility function plots

Both dashboards include:

• Alternative filter / ranking checklist (doubles as a selection panel)
• Highlight panel — assign any CSS colour to any alternative
• Weight sliders with automatic normalisation

Plots

Standalone Plotly figure builders — usable outside of a dashboard:

class / function	description
GaiaPlanePlot	Biplot of PROMETHEE GAIA projected coordinates
ParallelPlot	Parallel coordinates for raw performance tables or utility space
MAUTComparisonPlot	Criterion-wise weighted utility difference between two alternatives
build_maut_utility_plots	List of piecewise-linear utility function curves, one per criterion

All plot classes expose a build_figure() method returning a plotly.graph_objects.Figure.

from mcda_viz import GaiaPlanePlot
from promethee import PrometheeGAIA

gaia = PrometheeGAIA(performance_table=table, parameters=params)

fig = GaiaPlanePlot(
    gaia,
    visible_alternatives=["A", "B", "C"],   # optional subset
    highlight={"A": "red"},                  # optional colour map
).build_figure()   # plotly Figure — display in notebook or export

Components (building blocks for custom dashboards)

Each panel inside the built-in dashboards is a standalone DashboardComponent subclass. You can assemble them freely to build your own dashboard:

component	description
AlternativeSelectorComponent	A / B dropdown for pairwise comparison
AlternativeFilterComponent	Text-based alternative filter
ComparisonDescriptionComponent	Textual summary of A vs B weighted utility difference
CriterionPlotsComponent	Scrollable row of criterion utility function plots
GaiaPlanePlotComponent	GAIA plane wrapped as a Dash component
HighlightComponent	Colour picker — assign colours to alternatives
LegendComponent	Colour legend panel
ParallelPlotComponent	Parallel coordinates wrapped as a Dash component
ParametersComponent	Weight sliders and preference function selectors (PROMETHEE)
RankingChecklistComponent	Ranked list with checkboxes for filtering
RankingComponent	Read-only ranked list
StoreComponent	Invisible dcc.Store for shared dashboard state

Architecture

src/mcda_viz/
├── dashboards/
│   ├── core/          # BaseDashboard, DashboardComponent, DashboardState, shared IDs
│   ├── components/    # Reusable Dash components
│   ├── maut_analysis_dashboard.py
│   └── promethee_dashboard.py
├── plots/             # Pure Plotly figure builders
└── base/              # Shared plot utilities

DashboardState is a lightweight container (alternative names, criterion names, and optionally scores and ranking) instantiated once per dashboard and injected into every component that needs it.

BaseDashboard provides shared callbacks (highlight apply/reset, weight normalisation) and a Dash app factory. Subclass it to build a custom dashboard:

import dash_bootstrap_components as dbc
from dash import html
from mcda_viz import BaseDashboard, DashboardState, HighlightComponent, StoreComponent

class MyDashboard(BaseDashboard):
    def __init__(self, alternative_names, criteria_names):
        self.state = DashboardState(
            alternative_names=alternative_names,
            criteria_names=criteria_names,
        )
        self.app = self._create_app("My Dashboard")
        self.store_component = StoreComponent()
        self.highlight_component = HighlightComponent(self.state)
        self.app.layout = dbc.Container([
            self.store_component.layout(),
            self._header_row(),
            self.highlight_component.layout(),
        ])
        self._register_highlight_callback(self.app)

    def run(self, **kwargs):
        self.app.run(**kwargs)

Ecosystem

Part of the modular MCDA Python ecosystem built on mcda-core. See mcda-core for the full list of compatible libraries.

License

MIT — see LICENSE.