Dhaka Livability Index

A. Data sources

Amenities were collected from Google Places Aggregate API. Mobility was collected from Google Routes API. Air quality was collected from Google Air Quality API history lookup. No live APIs are called from this portfolio page; it uses a frozen local dataset.

B. Amenity collection

For each selected area, the system used the area center coordinate and a 1500m radius. For each amenity category, it collected mapped place counts. The count is mapped amenity availability from Google Places, not an official census.

C. Why the radius matters

The amenity model is intentionally a 1500m radius around a selected center point for each area. The measured area of that circle is:

The index compares amenity density inside this circle, so a compact dense area can score better than a larger more spread out area. This is intentional: the current model is measuring nearby access, not total area level inventory.

A future version could compare multiple radius options such as 1500m, 2500m, and 3000m.

D. Amenity category mapping

E. How individual category scores are calculated

For every area and every amenity category, the system starts with the raw mapped count from Google Places Aggregate API. Example: Area: Banani; Category: restaurant; Radius: 1500m; Raw count: number of mapped restaurants found within 1500m.

The raw count is converted into density so every area can be compared fairly.

Google Places counts can be very high in dense areas. log1p reduces the impact of extreme counts so one category does not dominate the score only because the raw count is very large.

For each category, the area with the strongest transformed density gets 100. The area with the weakest transformed density gets 0. Other areas receive scores between 0 and 100.

If Banani has the highest transformed restaurant density among the selected 10 areas, Banani gets restaurant_score = 100. If Bashundhara has the lowest transformed restaurant density, Bashundhara gets restaurant_score = 0. This does not mean Bashundhara has no restaurants. It only means it is weakest for restaurant density within this selected comparison set.

F. Topic score calculation

When a topic has multiple categories, the topic score is the average of those category scores.

• Healthcare: average of hospital_score and pharmacy_score
• Finance: average of bank_score and atm_score
• Food and Daily Needs: average of restaurant_score and supermarket_score
• Education: school_score
• Transport Access: bus_station_score
• Public Space Access: park_score

G. Essential Amenities calculation

Essential Amenities is calculated from Healthcare, Education, Finance, Food and Daily Needs, and Transport Access.

H. Mobility scoring

The mobility score uses an evening peak mobility sample. Each area is routed to every other selected area. Self routes are excluded. For 10 areas, this creates 90 route pairs.

The system calculates average travel time for each area. Lower average travel time gives a higher mobility score. This avoids the old bias where Motijheel or Gulshan scored well because they were also fixed destinations.

I. Air quality scoring

The air quality score uses 7 days of hourly Google Universal AQI data. Google Universal AQI is higher is better. Median AQI is used for scoring. PM2.5 and PM10 are shown as diagnostics, but they are not the main score.

Air quality is center point based, so it represents the selected coordinate, not every street inside the neighborhood.

J. Overall formula

Data Confidence reflects whether the required data components were available for the area.

K. Limitations

1. Scores are relative to the selected 10 areas.
2. Google Places counts are mapped availability, not an official census.
3. The 1500m radius measures local amenity access around a selected center point.
4. It does not cover the full boundary of large or irregular neighborhoods.
5. Results for larger or more spread out areas (for example, Bashundhara) may be sensitive to the chosen center coordinate and radius.
6. Scores should be read as comparative signals, not absolute judgments.
7. Amenity quality is not measured.
8. Public Space Access uses parks as a proxy.
9. Air quality is center point based.
10. Mobility reflects one evening peak collection window.
11. This is a comparative experimental index, not an official urban planning model.

All scores are relative to the selected 10 areas. A score of 100 means strongest within this dataset, not perfect. A score of 0 means weakest within this dataset, not absent.