toolathon123/project_20260813_012304_3c5740b5-gov-efficiency

Oceania Government – Weekly Efficiency Dataset

Weekly public-service request records for a Canberra-based Australian government agency (~180 staff), published to Hugging Face Hub for centralised, versioned efficiency analytics. This is the dataset for reference week 2026-W33 (request dates 2026-08-10 → 2026-08-13).

• Rows: 769
• Repository: https://huggingface.co/datasets/toolathon123/project_20260813_012304_3c5740b5-gov-efficiency
• Cadence: refreshed weekly (week_number identifies the ISO-8601 week)

Data fields

Field	Type	Description
request_id	string	Unique request identifier, e.g. REQ-2026-0001
region	string	ISO 3166-2 standard region code, e.g. AU-NSW, AU-VIC
service_type	string	Service category (enquiry, permit, complaint, …)
request_date	date	Date the request was lodged (YYYY-MM-DD)
response_time_hours	float	Time from lodgement to response in hours
resolution_status	string	resolved, closed, pending, in_progress, escalated, rejected, cancelled
channel	string	Intake channel (portal, email, phone, …)
staff_id	string	Case owner (STF-001 … STF-180)
week_number	int	ISO-8601 week number of request_date
year	int	ISO year of request_date

Cleaning & standardisation

1. Drop invalid response times – removed 31 records where response_time_hours was empty (NaN) or negative (see raw-vs-clean counts below).
2. Unify region to Oceania standard codes – mapped Australian state/territory names (and mixed Excel/internal-DB alias spellings such as NSW, VIC, QLD, New South Wales, Australian Capital Territory) to ISO 3166-2 codes (AU-NSW, AU-VIC, AU-QLD, AU-WA, AU-SA, AU-TAS, AU-ACT, AU-NT).
3. Add week_number – derived from request_date via ISO-8601 (datetime.date.isocalendar()), constant 33 for this week.

Raw simulated intake for this week was 800 records; 769 records remain after cleaning.

Statistical summary – week 2026-W33

Metric	Value
Total cleaned records	769
Average response time	29.81 hours
Median response time	21.0 hours
On-time completion rate (resolved / closed)	70.48% (542 of 769)

Resolution status distribution

Status	Count
resolved	334
closed	208
pending	97
in_progress	75
escalated	32
rejected	17
cancelled	6

Average response time by region (hours)

| Region | Mean (h) | Count | |---|---| | AU-ACT | 26.77 | 68 | | AU-NSW | 29.79 | 140 | | AU-NT | 35.08 | 56 | | AU-QLD | 26.15 | 143 | | AU-SA | 28.56 | 66 | | AU-TAS | 33.88 | 33 | | AU-VIC | 33.03 | 176 | | AU-WA | 27.77 | 87 |

Requests by channel

online_portal (218), email (189), phone (150), in_person (89), mobile_app (57), mail (38), chatbot (28)

Quick start

from datasets import load_dataset

ds = load_dataset("toolathon123/project_20260813_012304_3c5740b5-gov-efficiency")["train"]
print(len(ds))                          # 769
avg_response = ds.to_pandas()["response_time_hours"].mean()
on_time = ds.to_pandas()["resolution_status"].isin(["resolved", "closed"]).mean()

Pipeline & reproducibility

pipeline_build_dataset.py in this repo reproduces the full generate → clean → push pipeline (fixed random seed 42). The raw (pre-clean) records are kept locally and referenced in the script; only the processed dataset is published here for analysis.