CoolFace
Datasetpublic

THULab/africa_telecom_call_detail_records_nigeria

Africa Telecom Call Detail Records Nigeria TsFile This dataset is a TsFile conversion of electricsheepafrica/africa-synth-telecom-call-detail-records-nigeria, a synthetic Nigerian telecom call detail record dataset. Modalities: Time-series. Source Dataset Original dataset: electricsheepafrica/africa-synth-telecom-call-detail-records-nigeria Source files: call_detail_records.parquet and a CSV copy Source rows: 1,000,000 Source columns: 13 License: gpl… See the full description on the dataset page: https://huggingface.co/datasets/THULab/africa_telecom_call_detail_records_nigeria.

sourceHugging Faceotherupdated 1mo agoView on Hugging Face
0likes51downloads
Dataset Card

Africa Telecom Call Detail Records Nigeria TsFile

This dataset is a TsFile conversion of `electricsheepafrica/africa-synth-telecom-call-detail-records-nigeria`, a synthetic Nigerian telecom call detail record dataset.

Modalities: Time-series.

Source Dataset

Converted Data

  • —TsFile path: africa_telecom_call_detail_records_nigeria.tsfile
  • —TsFile table: call_detail_records
  • —Rows: 1,000,000
  • —Time precision: milliseconds
  • —Time range: 2025-09-01 00:00:04 UTC to 2025-10-01 23:59:55 UTC
  • —Source timestamp type: timestamp[ns]
  • —Timestamp precision check: all source timestamps are second-aligned, so converting to millisecond Time does not lose sub-millisecond information.

Schema Mapping

timestamp is converted to the TsFile Time column as epoch milliseconds and is not retained as a duplicate field.

TAG columns:

  • —operator
  • —city
  • —tower_id
  • —event_rank

FIELD columns:

  • —cdr_id
  • —calling_number
  • —called_number
  • —network_type
  • —call_type
  • —call_status
  • —duration_seconds
  • —sms_count
  • —data_volume_mb

Conversion Notes

tower_id is the primary telecom site/device dimension. operator and city are kept as TAG columns for direct filtering. cdr_id, calling_number, and called_number are unique or near-unique identifiers, so they remain FIELD columns rather than high-cardinality TAG columns.

The generated event_rank TAG preserves all CDR events without modifying millisecond Time. It is the duplicate order within (operator, city, tower_id, Time): 999,997 rows have event_rank=0, and 3 rows have event_rank=1. The final (operator, city, tower_id, event_rank, Time) key has zero duplicates.

Minimal Read Example

python
# Read africa_telecom_call_detail_records_nigeria.tsfile with the Apache TsFile SDK.
# Example logical filter:
#   operator = 'MTN' AND city = 'Lagos' AND tower_id = '<tower id>'