electricsheepafrica/africa-water-point-locations-based-on-2012-survey

---
annotations_creators:
- no-annotation
language_creators:
- found
language:
- en
license: cc-by-4.0
multilinguality:
- monolingual
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- other
task_ids: []
tags:
- africa
- humanitarian
- hdx
- electric-sheep-africa
- geodata
- water-sanitation-and-hygiene-wash
- sle
pretty_name: "Sierra Leone water point locations based on 2012 survey"
dataset_info:
splits:
- name: train
num_examples: 23076
- name: test
num_examples: 5769
---

# Sierra Leone water point locations based on 2012 survey

**Publisher:** MapAction · **Source:** [HDX](https://data.humdata.org/dataset/water-point-locations-based-on-2012-survey) · **License:** `cc-by-igo` · **Updated:** 2023-03-03

---

## Abstract

Country-wide dataset of water sources, including type, installer, quality.

Each row in this dataset represents subnational administrative unit observations. Data was last updated on HDX on 2023-03-03. Geographic scope: **SLE**.

*Curated into ML-ready Parquet format by [Electric Sheep Africa](https://huggingface.co/electricsheepafrica).*

---

## Dataset Characteristics

| | |
|---|---|
| **Domain** | Water, sanitation and hygiene (wash) |
| **Unit of observation** | Subnational administrative unit observations |
| **Rows (total)** | 28,845 |
| **Columns** | 39 (11 numeric, 28 categorical, 0 datetime) |
| **Train split** | 23,076 rows |
| **Test split** | 5,769 rows |
| **Geographic scope** | SLE |
| **Publisher** | MapAction |
| **HDX last updated** | 2023-03-03 |

---

## Variables

**Geographic** — `x` (range -13.2929–-10.2794), `y` (range 6.9326–9.9946), `wtype` (Pump on hand-dug well, Protected Well (no pump), Standpipe or Tapstand), `s_wtype` (local well, hand dug well, protected well), `pmptype` (Bucket, India Mark, Kardia) and 5 others.

**Temporal** — `subdate` (#VALUE!, 02-07-2012 8:28:43 EST, 02-09-2012 8:07:55 EST), `season`.

**Demographic** — `age` (range 1961.0–2012.0), `manager`.

**Identifier / Metadata** — `code`, `name`, `eacode` (range 11010101.0–42081019.0), `esa_source`, `esa_processed`.

**Other** — `instance` (f1b0fb34-99ec-4f01-a952-94b376e56abe, 7e37e787-b8b1-498a-9e4d-a41e6f3a47a0, d03858b2-1d8e-40e5-b78c-a204edbc1a7e), `sub` (user 14, user 16, user 24), `funct` (Yes- functional, No- broken down, Yes- but partly damaged), `tapno` (range 1.0–45.0), `dmg_all` (Pump damaged, Tap head(s) broken or missing, Pipe(s) blocked or disconnected) and 15 others.

---

## Quick Start

```python
from datasets import load_dataset

ds = load_dataset("electricsheepafrica/africa-water-point-locations-based-on-2012-survey")
train = ds["train"].to_pandas()
test = ds["test"].to_pandas()

print(train.shape)
train.head()
```

---

## Schema

| Column | Type | Null % | Range / Sample Values |
|---|---|---|---|
| `x` | float64 | 0.0% | -13.2929 – -10.2794 (mean -11.9018) |
| `y` | float64 | 0.0% | 6.9326 – 9.9946 (mean 8.3661) |
| `instance` | object | 0.0% | f1b0fb34-99ec-4f01-a952-94b376e56abe, 7e37e787-b8b1-498a-9e4d-a41e6f3a47a0, d03858b2-1d8e-40e5-b78c-a204edbc1a7e |
| `subdate` | object | 0.0% | #VALUE!, 02-07-2012 8:28:43 EST, 02-09-2012 8:07:55 EST |
| `sub` | object | 0.0% | user 14, user 16, user 24 |
| `funct` | object | 0.0% | Yes- functional, No- broken down, Yes- but partly damaged |
| `wtype` | object | 0.0% | Pump on hand-dug well, Protected Well (no pump), Standpipe or Tapstand |
| `s_wtype` | object | 77.1% | local well, hand dug well, protected well |
| `tapno` | float64 | 72.5% | 1.0 – 45.0 (mean 1.2777) |
| `pmptype` | object | 30.0% | Bucket, India Mark, Kardia |
| `dmg_all` | object | 67.9% | Pump damaged, Tap head(s) broken or missing, Pipe(s) blocked or disconnected |
| `age` | float64 | 19.7% | 1961.0 – 2012.0 (mean 2003.1499) |
| `install` | object | 0.0% | Private, United Nations (UN), Other |
| `install_s` | object | 0.7% | Private Person, United Nations (UN), Religious Group |
| `used` | object | 5.1% | |
| `season` | object | 5.1% | |
| `qual` | object | 11.8% | |
| `chlorine` | object | 5.1% | |
| `manager` | object | 0.2% | |
| `money` | object | 5.2% | |
| `mechanic` | object | 0.0% | |
| `parts` | object | 0.0% | |
| `lat` | float64 | 0.0% | 6.9326 – 9.9946 (mean 8.3661) |
| `lon` | float64 | 0.0% | -13.2929 – -10.2794 (mean -11.9018) |
| `alt` | float64 | 0.0% | -1227.2 – 3899.0 (mean 171.1209) |
| `code` | object | 0.0% | |
| `name` | object | 0.0% | |
| `photo` | object | 0.0% | |
| `eacode` | int64 | 0.0% | 11010101.0 – 42081019.0 (mean 24997284.0522) |
| `province` | object | 0.0% | |
| `district` | object | 0.0% | |
| `localcounc` | object | 0.0% | |
| `chiefdom` | object | 0.0% | |
| `section` | object | 0.0% | |
| `wpaccess` | float64 | 0.0% | 0.0 – 6.0 (mean 0.6877) |
| `wpns_acc` | float64 | 0.0% | 0.0 – 6.0 (mean 0.4047) |
| `tw` | float64 | 0.0% | 1.0 – 9.0 (mean 1.057) |
| `esa_source` | object | 0.0% | |
| `esa_processed` | object | 0.0% | |

---

## Numeric Summary

| Column | Min | Max | Mean | Median |
|---|---|---|---|---|
| `x` | -13.2929 | -10.2794 | -11.9018 | -11.7519 |
| `y` | 6.9326 | 9.9946 | 8.3661 | 8.3253 |
| `tapno` | 1.0 | 45.0 | 1.2777 | 1.0 |
| `age` | 1961.0 | 2012.0 | 2003.1499 | 2006.0 |
| `lat` | 6.9326 | 9.9946 | 8.3661 | 8.3253 |
| `lon` | -13.2929 | -10.2794 | -11.9018 | -11.7519 |
| `alt` | -1227.2 | 3899.0 | 171.1209 | 124.9 |
| `eacode` | 11010101.0 | 42081019.0 | 24997284.0522 | 24071005.0 |
| `wpaccess` | 0.0 | 6.0 | 0.6877 | 1.0 |
| `wpns_acc` | 0.0 | 6.0 | 0.4047 | 0.0 |
| `tw` | 1.0 | 9.0 | 1.057 | 1.0 |

---

## 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`. 12 column(s) with >80% missing values were removed: `dmg_well`, `dmg_apron`, `dmg_pmp`, `dmg_pipe`, `dmg_tap`, `dmg_concre`.... 2 column(s) were cast from string to numeric or datetime based on parse-success rate (>85% threshold). 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 MapAction and has not been independently validated by ESA.
- Automated cleaning cannot correct for misreported values, definitional inconsistencies, or sampling bias in the original collection.
- The following columns have >20% missing values and should be treated with caution in modelling: `s_wtype`, `tapno`, `pmptype`, `dmg_all`.
- Refer to the [original HDX dataset page](https://data.humdata.org/dataset/water-point-locations-based-on-2012-survey) for the publisher's own methodology notes and caveats.

---

## Citation

```bibtex
@dataset{hdx_africa_water_point_locations_based_on_2012_survey,
title = {Sierra Leone water point locations based on 2012 survey},
author = {MapAction},
year = {2023},
url = {https://data.humdata.org/dataset/water-point-locations-based-on-2012-survey},
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.*