---
license: cc-by-nc-nd-4.0
task_categories:
- time-series-forecasting
language:
- en
tags:
- tsfile
- timeseries
- time-series
- macroeconomics
- forecasting
- economics
- finance
- format:tsfile
pretty_name: MacroEconHorizon (TsFile)
size_categories:
- 1M<n<10M
---

# MacroEconHorizon (TsFile)

Apache TsFile version of [`bkoyuncu/MacroEcon`](https://huggingface.co/datasets/bkoyuncu/MacroEcon).

## Overview

MacroEconHorizon is a curated collection of macroeconomic time series spanning 63
countries (plus the euro area and South Africa, 64 country files in total) from
1970 to 2023. It is designed to support nowcasting, forecasting, and scenario
analysis for machine-learning researchers and economic policy makers. Indicators
cover GDP, inflation, unemployment rates, commodity prices, financial market
variables, and monetary policy metrics, curated by the Bank for International
Settlements (BIS).

Each variable is identified by a code, with a frequency (D=Daily, M=Monthly,
Q=Quarterly) and a transformation (`value`, `_delta`, `_log`, `_delta_log`).
Frequencies are standardized to a daily grid via forward-fill.

- **Countries:** 64 (63 countries + euro area)
- **Time span:** 1970 to 2023 (daily grid)
- **Rows:** ~1.3 million across all country panels
- **Indicators:** GDP, inflation, unemployment, commodity prices, financial
  variables, monetary policy metrics

## Schema (TsFile structure)

Because the feature columns are country-specific (the indicator set does not
overlap across countries), each country is converted to its own TsFile / table
`macroecon_<cc>` rather than a single sparse union table.

- **Time** (INT64, milliseconds) — the daily date (source `Unnamed: 0` column).
- **`Q_...` / `D_...` / `M_...` / `A_...`** (FLOAT) — the economic indicators.
  Column names encode frequency, code, country, series row, and transformation
  (e.g. `Q_FXSP_AE_row_0_value_delta_log`); dots/special characters are replaced
  with underscores.

Each country is a single device; its indicators are numeric FIELD columns.
Missing values (forward-fill does not reach the start of every series) are stored
as null.

## Usage

Install the Apache TsFile Python SDK (`pip install tsfile`) and read a converted file:

```python
from pathlib import Path
from tsfile import TsFileReader

path = Path("macroecon_ae.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())
```

## Source & license

- Original dataset: https://huggingface.co/datasets/bkoyuncu/MacroEcon
- Author / publisher: bkoyuncu
- Curated by: Bank for International Settlements (BIS)
- License: CC BY-NC-ND 4.0