Add README.md

README.md

@@ -1,48 +1,233 @@
 ---
+annotations_creators:
+- no-annotation
+language_creators:
+- found
+language:
+- en
+license: cc-by-sa-4.0
+multilinguality:
+- monolingual
+size_categories:
+- 1K<n<10K
+source_datasets:
+- original
+task_categories:
+- tabular-classification
+- tabular-regression
+- other
+task_ids: []
+tags:
+- africa
+- humanitarian
+- hdx
+- electric-sheep-africa
+- education
+- health-facilities
+- transportation
+- sdn
+pretty_name: "Sudan - Accessibility Indicators"
 dataset_info:
-  features:
-  - name: name
-    dtype: string
-  - name: id
-    dtype: string
-  - name: country
-    dtype: string
-  - name: admin_level
-    dtype: string
-  - name: category
-    dtype: string
-  - name: range_type
-    dtype: string
-  - name: range
-    dtype: int64
-  - name: population_type
-    dtype: string
-  - name: population
-    dtype: int64
-  - name: population_share
-    dtype: float64
-  - name: population_interval
-    dtype: int64
-  - name: population_interval_share
-    dtype: float64
-  - name: esa_source
-    dtype: string
-  - name: esa_processed
-    dtype: string
   splits:
-  - name: train
-    num_bytes: 454741
-    num_examples: 2972
-  - name: test
-    num_bytes: 113921
-    num_examples: 744
-  download_size: 118977
-  dataset_size: 568662
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-  - split: test
-    path: data/test-*
+    - name: train
+      num_examples: 2972
+    - name: test
+      num_examples: 743
 ---
+
+# Sudan - Accessibility Indicators
+
+**Publisher:** HeiGIT (Heidelberg Institute for Geoinformation Technology) · **Source:** [HDX](https://data.humdata.org/dataset/sudan-accessibility-indicators) · **License:** `cc-by-sa` · **Updated:** 2026-02-27
+
+---
+
+## Abstract
+
+This dataset provides insights into spatial accessibility to healthcare and
+education services across Sudan. It has been created using free and
+open tools such as [openrouteservice](https://openrouteservice.org/) and open
+data sources, primarily [OpenStreetMap](https://www.openstreetmap.org/) (OSM).
+
+
+To assess accessibility to education and healthcare, we use travel-time 
+isochrones—polygons representing areas reachable within a given time or distance by 
+car. We overlay these isochrones with [WorldPop](https://www.worldpop.org/) population 
+data, which provides 100m-resolution estimates. This allows us to calculate the 
+population within time intervals from 10 to 120 minutes away from hospital services and 
+distance intervals from 5 to 50 km away from schools. The unit of analysis is defined 
+by [geoboundaries](https://www.geoboundaries.org/) country borders, and where available 
+we also summarise results at finer administrative levels (ADM 1–4).
+
+
+Data Structure:
+
+- **name**: Region or country name.
+- **iso**: ISO3 country code.
+- **id**: Unique identifier for the administrative unit.
+- **country**: ISO3 country code.
+- **admin_level**: Administrative level of the unit.
+- **category**: Service category — `education`, `hospitals` or
+  `primary_healthcare`.
+- **range_type**: Method used for the catchment zone — `distance` or `time`.
+- **range**: Distance (in meters) or Time away (in seconds) from schools used
+  to generate the polygon.
+- **population**: Total population within the specified range.
+- **school_age_population**: Number of school-age individuals within the range.
+- **school_age_population_share**: Cumulative percentage of school-age
+  population.
+- **school_age_population_interval**: Incremental school-age population added
+  in the current distance band.
+- **school_age_population_interval_share**: Proportion of new school-age
+  population in the current interval.
+- **population_share**: Cumulative percentage of total population.
+- **population_interval**: Incremental population added in the current distance
+  band.
+- **population_interval_share**: Share of the total population represented by
+  the current interval.
+
+This dataset is one of many [HeiGIT exports on HDX](https://data.humdata.org/organization/heidelberg-institute-for-geoinformation-technology).
+See the [HeiGIT](https://heigit.org/) website for more information.
+
+We are looking forward to hearing about your use-case! Feel free to reach out
+to us and tell us about your research at
+[communications@heigit.org](mailto:communications@heigit.org) – we would be
+happy to amplify your work.
+
+References:
+
+- [Geldsetzer, P., Reinmuth, M., Ouma, P. O., Lautenbach, S. et al. (2020)](https://www.thelancet.com/journals/lanhl/article/PIIS2666-7568(20)30010-6/fulltext)
+- [Petricola, S., Reinmuth, M., Lautenbach, S. et al. (2022)](https://ij-healthgeographics.biomedcentral.com/articles/10.1186/s12942-022-00315-2)
+- [Klipper, I. G., Zipf, A., and Lautenbach, S. (2021)](https://agile-giss.copernicus.org/articles/2/4/2021/)
+- [Ruiz Sánchez, R., Reinmuth, M., Albornoz, C., Lautenbach, S., and Zipf, A. (2025)](https://agile-giss.copernicus.org/articles/6/10/2025/)
+
+Further Information:
+
+- [Open Access Lens](https://giscience.github.io/open-access-lens/#/)
+
+**Limitations**:
+
+* **OSM Completeness**: This analysis relies on OpenStreetMap (OSM) data. While OSM is
+the most complete open map of the world, data quality varies significantly by region.
+In areas with unmapped roads or facilities, accessibility may be underestimated.
+
+* **Population Estimates**: Population counts are derived from WorldPop top-down
+estimates (constrained). These are statistical models based on census projections and
+satellite imagery, not direct census counts, and may contain inaccuracies at the local
+pixel level.
+
+* **Travel Time Assumptions**: Isochrones are calculated using standard vehicle speeds
+for different road types. These models do not account for real-time traffic, seasonal
+weather conditions (e.g., flooding), or road surface degradation.
+
+* **Boundary Precision**: Administrative boundaries are sourced from geoBoundaries.
+These may differ slightly from official government demarcations or other schemas.
+
+Each row in this dataset represents country-level aggregates. Data was last updated on HDX on 2026-02-27. Geographic scope: **SDN**.
+
+*Curated into ML-ready Parquet format by [Electric Sheep Africa](https://huggingface.co/electricsheepafrica).*
+
+---
+
+## Dataset Characteristics
+
+| | |
+|---|---|
+| **Domain** | Public health |
+| **Unit of observation** | Country-level aggregates |
+| **Rows (total)** | 3,716 |
+| **Columns** | 14 (5 numeric, 9 categorical, 0 datetime) |
+| **Train split** | 2,972 rows |
+| **Test split** | 743 rows |
+| **Geographic scope** | SDN |
+| **Publisher** | HeiGIT (Heidelberg Institute for Geoinformation Technology) |
+| **HDX last updated** | 2026-02-27 |
+
+---
+
+## Variables
+
+**Geographic** — `country` (SDN), `admin_level` (ADM2, ADM1, ADM0), `category` (education), `range_type` (DISTANCE), `population_type` (school_age, total) and 4 others.
+
+**Identifier / Metadata** — `name` (Khartoum, Ar Rahad, Sennar), `id` (16777658B19857105557319, 16777658B27261158380636, 16777658B23897511285986), `esa_source` (HDX), `esa_processed` (2026-04-27).
+
+**Other** — `range` (range 5000.0–50000.0).
+
+---
+
+## Quick Start
+
+```python
+from datasets import load_dataset
+
+ds    = load_dataset("electricsheepafrica/africa-education-sudan")
+train = ds["train"].to_pandas()
+test  = ds["test"].to_pandas()
+
+print(train.shape)
+train.head()
+```
+
+---
+
+## Schema
+
+| Column | Type | Null % | Range / Sample Values |
+|---|---|---|---|
+| `name` | object | 0.0% | Khartoum, Ar Rahad, Sennar |
+| `id` | object | 0.0% | 16777658B19857105557319, 16777658B27261158380636, 16777658B23897511285986 |
+| `country` | object | 0.0% | SDN |
+| `admin_level` | object | 0.0% | ADM2, ADM1, ADM0 |
+| `category` | object | 0.0% | education |
+| `range_type` | object | 0.0% | DISTANCE |
+| `range` | int64 | 0.0% | 5000.0 – 50000.0 (mean 28100.1076) |
+| `population_type` | object | 0.0% | school_age, total |
+| `population` | int64 | 0.0% | 0.0 – 21624249.0 (mean 209500.3122) |
+| `population_share` | float64 | 0.0% | 0.0 – 100.0 (mean 31.0481) |
+| `population_interval` | int64 | 0.0% | 0.0 – 11360923.0 (mean 25642.4244) |
+| `population_interval_share` | float64 | 0.0% | 0.0 – 97.04 (mean 4.0914) |
+| `esa_source` | object | 0.0% | HDX |
+| `esa_processed` | object | 0.0% | 2026-04-27 |
+
+---
+
+## Numeric Summary
+
+| Column | Min | Max | Mean | Median |
+|---|---|---|---|---|
+| `range` | 5000.0 | 50000.0 | 28100.1076 | 30000.0 |
+| `population` | 0.0 | 21624249.0 | 209500.3122 | 28046.5 |
+| `population_share` | 0.0 | 100.0 | 31.0481 | 22.79 |
+| `population_interval` | 0.0 | 11360923.0 | 25642.4244 | 2013.0 |
+| `population_interval_share` | 0.0 | 97.04 | 4.0914 | 1.47 |
+
+---
+
+## Curation
+
+Raw data was downloaded from HDX via the CKAN API and converted to Parquet. Column names were lowercased and standardised to snake_case. Common missing-value markers (`N/A`, `null`, `none`, `-`, `unknown`, `no data`, `#N/A`) were unified to `NaN`. 1 column(s) with >80% missing values were removed: `iso`. The dataset was split 80/20 into train and test partitions using a fixed random seed (42) and saved as Snappy-compressed Parquet.
+
+---
+
+## Limitations
+
+- Data originates from HeiGIT (Heidelberg Institute for Geoinformation Technology) and has not been independently validated by ESA.
+- Automated cleaning cannot correct for misreported values, definitional inconsistencies, or sampling bias in the original collection.
+- Refer to the [original HDX dataset page](https://data.humdata.org/dataset/sudan-accessibility-indicators) for the publisher's own methodology notes and caveats.
+
+---
+
+## Citation
+
+```bibtex
+@dataset{hdx_africa_education_sudan,
+  title     = {Sudan - Accessibility Indicators},
+  author    = {HeiGIT (Heidelberg Institute for Geoinformation Technology)},
+  year      = {2026},
+  url       = {https://data.humdata.org/dataset/sudan-accessibility-indicators},
+  note      = {Repackaged for machine learning by Electric Sheep Africa (https://huggingface.co/electricsheepafrica)}
+}
+```
+
+---
+
+*[Electric Sheep Africa](https://huggingface.co/electricsheepafrica) — Africa's ML dataset infrastructure. Lagos, Nigeria.*