MGWRVisualizer 🕸️🗺

Objective

Learning Multiscale Geographically Weighted Regression (MGWR) can prove difficult and unaccesible. This website + clients aim to address this problem by creating easy-to-understand interactive visualizations of MGWR properties.

Deployed Test Site

Project Status

• Python client - ✔️
  • visualize with local server ✔️
  • save results to upload to web client ✔️
• Web client - 🟨 functional
  • pre-built datasets to visualize - ❌
  • upload dataset to visualize - ❌

Current Features

• Spatial viz of covariate bandwith
• Spatial viz of attribute bandwith
• Dynamic chart of signifigant parameters at 95th confidence
• Model results
• Diagnostic information

Work in Progress

• viz model fit based on AIC versus bandwith

Python Client

The Python client can do 2 main things

• Run a local mgwrvisualizer webclient server
• Export formatted MGWR result objects for later visualization

Install

pip install mgwrvisualizer

Usage

See examples for full example Jupyter notebooks

Running local server

from mgwrvisualizer import MGWRVisualizer

data_df = pd.DatFrame ...  # non-spatial dataframe
geodata_df = gpd.GeoDataFrame ...  # spatial dataframe
mgwr_results = MGWR(coords, _y, _X, mgwr_selector).fit()

viz = MGWRVisualizer(mgwr_results, data_df, geodata_df, merge_key="AreaKey")

viz.run()  # run server

>> browser opens to local server

Saving results

You can save a formatted version of the MGWR model results. You can then upload this file to the webclient for later visualization (in-progress)

from mgwrvisualizer import MGWRVisualizer

data_df = pd.DatFrame ...  # non-spatial dataframe
geodata_df = gpd.GeoDataFrame ...  # spatial dataframe
mgwr_results = MGWR(coords, _y, _X, mgwr_selector).fit()

viz = MGWRVisualizer(mgwr_results, data_df, geodata_df, merge_key="AreaKey")

viz.save_results("mgwr-results-file.json")