India Open Government Data — Bharat Forecasting Datasets
Collection
Research and non-commercial use (CC BY-NC-SA 4.0). Commercial: surya@judicialmind.ai / hello@judicialmind.ai • 1 item • Updated
Datasets:
You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This dataset is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 licence (CC BY-NC-SA 4.0): research, academic, personal and other non-commercial use is permitted with attribution; adaptations must be shared under the same licence; commercial use is not permitted. The original public source data remains under the Government Open Data License - India (GODL-India). Commercial use: contact surya@judicialmind.ai or hello@judicialmind.ai.
Log in or Sign Up to review the conditions and access this dataset content.
BharatTS-Corpus
Licence: CC BY-NC-SA 4.0 — research and non-commercial use only; derivatives must be shared under the same licence. Commercial use: contact surya@judicialmind.ai or hello@judicialmind.ai
Access is approved automatically when the email you enter matches your Hugging Face account email and belongs to the organisation you name; other requests are reviewed manually.
1,799,285 time series (76,331,896 observed points) extracted from 8,461 Indian government tables (data.gov.in, NITI Aayog NDAP, other government portals) plus curated series (Agmarknet mandi prices, HMIS, crop yields, rainfall, reservoirs, PM-KISAN). Daily, weekly, monthly, quarterly and yearly series from 73 publishing ministries, departments and organisations, one row per series. Near-duplicates were removed (same frequency, start and z-normalised shape). This is the pre-training corpus of BharatTS-Large; its heldout series are BharatTS-Large's held-out evaluation set.
Quickstart
# --- datasets
from datasets import load_dataset
ds = load_dataset("judicialmind/BharatTS-Corpus", split="train") # gated: accept the terms, then `hf auth login`
print(ds.num_rows, ds.column_names)
# --- pandas: one series as a dated pandas Series
import pandas as pd
from huggingface_hub import hf_hub_download
path = hf_hub_download("judicialmind/BharatTS-Corpus", "series.parquet", repo_type="dataset")
df = pd.read_parquet(path, columns=["series_id", "ministry", "title", "entity", "measure", "freq",
"start_date", "values", "n_obs", "split"])
row = df[(df.freq == "M") & df.title.str.contains("rainfall", case=False)].iloc[0]
step = {"D": "D", "W": "7D", "M": "MS", "Q": "QS", "Y": "YS"}[row.freq]
s = pd.Series(row["values"], index=pd.date_range(row.start_date, periods=len(row["values"]), freq=step))
print(row.title, "|", row.entity, "|", row.measure)
print(s.tail(3))
# --- duckdb: counts without loading the file
import duckdb
print(duckdb.sql(f"SELECT freq, count(*) AS series, sum(n_obs) AS observations FROM '{path}' "
"GROUP BY freq ORDER BY series DESC").df())
Output: ds has 1,799,285 rows with columns series_id, uuid, source, ministry, sector, title, entity, measure, freq, start, start_date, n_obs, length, values, split. The first monthly rainfall series is Heavy Rainfall Events | India | Day3 / 72 Hours | I2831_6:
Last 3 months of that series
| Month | Value |
|---|---|
| 2024-07-01 | 70 |
| 2024-08-01 | 67 |
| 2024-09-01 | 76 |
Series and observations by frequency
| Frequency | Series | Observations |
|---|---|---|
| D (daily) | 701,894 | 51,258,159 |
| M (monthly) | 663,229 | 19,019,347 |
| Y (yearly) | 276,418 | 2,846,089 |
| Q (quarterly) | 157,337 | 3,200,571 |
| W (weekly) | 407 | 7,730 |
values can be passed straight to BharatTS-Large: BharatTSLarge.from_pretrained(...).predict([row["values"]], freq=row.freq).
Files
| file | content |
|---|---|
series.parquet |
all series, one row per series (zstd parquet, about 205 MB) |
Schema
| column | type | description |
|---|---|---|
series_id |
string | unique id of the series |
uuid |
string | source table: data.gov.in resource id (UUID), ndap-<dataset id> for NDAP datasets, special-<name> for the curated series (special-mandi, special-hmis, special-crop, special-rain, special-reservoirs, special-pmkisan), hmis-state-<portal>-<id> for tables from other government portals |
source |
string | ogd (data.gov.in), ndap, special (curated series), HMIS state reports (data.gov.in) (other government portals) |
ministry |
string | publishing ministry / department / organisation of the source table |
sector |
string | sector of the source table as labelled by the portal |
title |
string | title of the source table |
entity |
string | place and breakdown the series belongs to, ` |
measure |
string | measured column of the source table |
freq |
string | D daily, W weekly (weeks start on Monday), M monthly, Q quarterly, Y yearly |
start |
int64 | integer period index of the first value: D days since 1970-01-01, W weeks since Monday 1969-12-29, M year12+month-1, Q year4+quarter-1, Y year |
start_date |
date | first day of the first period |
values |
list | one value per period from start_date, oldest first; NaN = missing |
n_obs |
int32 | number of non-missing values |
length |
int32 | number of periods (= length of values) |
split |
string | train or heldout (series never used to train BharatTS-Large; its held-out evaluation set) |
PM-KISAN series (special-pmkisan) step by installment and are stored with freq = Q and a nominal start_date.
Coverage
| freq | series | observed points | first start date | median points per series |
|---|---|---|---|---|
| D | 701,894 | 51,258,159 | 1901-02-01 | 27 |
| M | 663,229 | 19,019,347 | 1950-01-01 | 24 |
| Y | 276,418 | 2,846,089 | 1950-01-01 | 9 |
| Q | 157,337 | 3,200,571 | 2000-04-01 | 21 |
| W | 407 | 7,730 | 2011-05-09 | 12 |
| source | series | source tables | observed points |
|---|---|---|---|
| ogd (data.gov.in) | 1,119,230 | 8,073 | 28,176,105 |
| special (curated) | 402,370 | 6 | 22,588,077 |
| HMIS state reports (data.gov.in) | 186,171 | 11 | 7,521,866 |
| ndap | 91,514 | 371 | 18,045,848 |
Splits: train 1,709,420 series (72,479,022 points), heldout 89,865 series (3,852,874 points). India: national, state, district and sub-district series; most recent periods in 2026. Every series has at least 8 observed points; series hold at most 2,048 values. Kisan Call Centre call-count series are not included (published with small-cell suppression in KCC-Pulse).
Source data
Source data: Open Government Data Platform India (data.gov.in) and NDAP (ndap.niti.gov.in), used under the Government Open Data License - India (GODL-India), https://data.gov.in/government-open-data-license-india. Aggregates only; no personal data. NDAP series: NITI Aayog National Data and Analytics Platform terms of use.
For data.gov.in and NDAP series the uuid column gives the data.gov.in resource id or NDAP dataset id; ministry names the publisher of every series.
Original data: published by the Indian ministries, departments and agencies named per series in the ministry column on the Open Government Data Platform India (data.gov.in) and, where marked, NITI Aayog's National Data and Analytics Platform (NDAP), under the Government Open Data License - India (GODL-India). The original public data remains under GODL-India; the CC BY-NC-SA 4.0 licence of this repository covers JudicialMind's compilation, derived fields, annotations, benchmarks and models.
Citation
@misc{judicialmind2026bharattscorpus,
title = {BharatTS-Corpus: 1.8 Million Time Series from Indian Open Government Data},
author = {JudicialMind},
year = {2026},
publisher = {Hugging Face},
howpublished = {\url{https://huggingface.co/datasets/judicialmind/BharatTS-Corpus}}
}
License
CC BY-NC-SA 4.0 (see LICENSE): research, academic, personal and other non-commercial use, with attribution to JudicialMind and to the original data publishers; adaptations must be shared under the same licence. Commercial use requires a separate licence.
Commercial use: contact surya@judicialmind.ai or hello@judicialmind.ai.
- Downloads last month
- 17