AM Abdullah Mutlu
TR
All projects

Live demo /2025 /Geospatial · Full-stack · AI

ChargeScout

Data-driven, explainable site selection for charging-network operators: draw an area, score the best spots for a new station, and export a report.

Live demo

How it works

  1. 1

    Draw the area

    Draw the region to analyze on the map (satellite / standard / light basemap).

  2. 2

    Generate candidates

    The app builds a hex-grid of candidate points inside it.

  3. 3

    Multi-criteria scoring

    Coverage gap, demand, roads, traffic, parking, fuel; factor influence via scikit-learn.

  4. 4

    Rank & report

    Show the best sites with an explainable score and export a detailed PDF report.

Problem

Deciding where to build a new charging station depends on many criteria at once — distance to existing stations, road network, traffic, parking, fuel and population. Operators needed a data-driven, visual and explainable tool for that decision.

Solution

An OpenLayers GIS interface (multiple basemaps: standard/satellite/light, drawing and measurement tools, icon layers) lets the user draw an area. A Python/FastAPI service generates hex-grid candidate points inside it and scores each on multiple criteria (coverage gap, demand/population, road access, traffic, parking, fuel). Geometry uses shapely; the model uses scikit-learn (RandomForest for factor influence, KMeans for distinct zones); data is pulled from OpenStreetMap (Overpass). Results appear as a ranked list + score breakdown + map markers, and export to a detailed PDF report.

Results

A fully working decision-support tool: on real Ankara data it ranks the best candidate sites in a drawn area with an explainable score and exports a PDF report for investment planning.