---
license: other
task_categories:
- tabular-classification
- tabular-regression
language:
- en
tags:
- africa
- electric-sheep-africa
- open-data
- metadata-backed
- technology-digital
- parquet
- tabular
- text
- telecom
- cell
- tower
- handover
- nigeria
- mobility
- tsfile
- modality:timeseries
- timeseries
- format:tsfile
pretty_name: Africa Synthetic Telecom Cell Tower Handover (Nigeria) — TsFile
size_categories:
- 1M<n<10M
configs:
- config_name: default
  data_files:
  - split: train
    path: africa_telecom_cell_tower_handover_nigeria.tsfile
modality:
- timeseries
- tabular
- text
---

# Africa Synthetic Telecom Cell Tower Handover — Nigeria (TsFile)

Apache TsFile version of [`electricsheepafrica/africa-synth-telecom-cell-tower-handover-data-nigeria`](https://huggingface.co/datasets/electricsheepafrica/africa-synth-telecom-cell-tower-handover-data-nigeria).

## Overview

A synthetic dataset of cell tower handover events for the Nigerian mobile
network. Each record describes one handover — a subscriber's connection moving
from one cell tower to another — including the source and destination towers and
cities, whether it succeeded, its duration, the signal strength before and after,
and whether a call or data session was active during the handover.

The data incorporates authentic Nigerian telecommunications characteristics:
major cities (Lagos, Abuja, Kano, Port Harcourt, Ibadan, and others), 2G/3G/4G/5G
network types, and West Africa Time (WAT, UTC+1) event timing.

- **Scale:** 1,000,000 handovers × 15 columns.
- **Time span:** 2025-09-01 to 2025-10-01, second resolution.
- **Cities:** 15 distinct cities (`from_city` / `to_city`).
- **Handover types:** intracity, intercity.
- **Network types:** 2G, 3G, 4G, 5G.

## Schema (TsFile structure)

This is an event/log table, not a per-entity time series. The high-cardinality
identifier columns are kept as fields; the device dimension is the set of
low-cardinality geo/category columns, which yields ~900 device groups
(from_city × to_city × handover_type × network_type), each a dense per-second
handover series.

- **Time** (INT64, milliseconds) — handover timestamp. The source `timestamp`
  is second-aligned, so the millisecond conversion is lossless. The source
  `timestamp` column is replaced by `Time` and not retained as a duplicate field.
- **from_city, to_city, handover_type, network_type** (TAG) — the device
  dimension. Query one device group with, e.g.,
  `WHERE from_city='Lagos' AND to_city='Lagos' AND handover_type='intracity' AND network_type='4G'`.
- **event_rank** (TAG, INT64) — a generated disambiguator (0..2) that
  distinguishes concurrent handovers sharing the same (from_city, to_city,
  handover_type, network_type, Time) without altering Time. It is derived only
  from the duplicate order within each device-second group.
- **handover_id, customer_id, from_tower_id, to_tower_id** (FIELD, STRING) —
  `handover_id` is unique per row, `customer_id` is near-unique (~946k distinct),
  and the tower ids are high-cardinality (~135k each), so all stay as fields
  rather than device keys.
- **handover_duration_ms, signal_strength_from_dbm, signal_strength_to_dbm**
  (FIELD, INT64) — handover duration and the source/destination signal strength
  in dBm.
- **handover_success, call_in_progress, data_session_active** (FIELD, BOOLEAN) —
  outcome and activity flags.

No columns were dropped and no rows were dropped: all 15 source columns and
1,000,000 rows are preserved.

## Usage

Read the `.tsfile` file with the Apache TsFile Java or Python SDK.

## Source & license

- Original dataset: https://huggingface.co/datasets/electricsheepafrica/africa-synth-telecom-cell-tower-handover-data-nigeria
- Author / publisher: electricsheepafrica (Hugging Face)
- License: the Hugging Face dataset card metadata declares `gpl`, while the
  source README body states "MIT License - For educational and research
  purposes". Please defer to the original dataset for the authoritative terms.