Importing data¶

RealTimeModel accepts forecast_evaluation.ForecastData and its subclasses. Choose the container according to how often forecasts are issued:

• Use NowcastData when forecasts may be issued several times within one target period, such as weekly or monthly vintages for quarterly GDP.
• Use ForecastData when each target period has one forecast date, such as one quarterly forecast for each quarter.

The generated real-time data contains monthly vintages, so this example uses NowcastData:

import forecast_evaluation as fe
import forecast_realtime as rt

sample_data = rt.generate_synthetic_data(
    N=2,
    first_period="2015-01-31",
    endpoint="2024-12-31",
)
print(sample_data.head().to_string(index=False))

forecast_data = fe.NowcastData(outturns_data=sample_data)
rt_model = rt.RealTimeModel(data=forecast_data, models=models)

The generated data begins:

    date frequency  variable      value vintage_date metric
2015-01-31         M monthly_1 101.577869   2024-01-31 levels
2015-01-31         M monthly_2 101.256092   2024-01-31 levels
2015-02-28         M monthly_1 101.293703   2024-01-31 levels
2015-02-28         M monthly_1  -0.002798   2024-01-31    pop
2015-02-28         M monthly_2  98.456195   2024-01-31 levels

Outturn schema¶

The outturn DataFrame uses long-form rows with these columns:

date identifies the period measured; vintage_date identifies when the value became available. Use another row with a later vintage_date to record a revision. At each forecast vintage, RealTimeModel uses the latest value available by that date. Complete snapshots are not required.

Use levels for raw values, pop for period-on-period percentage growth, and yoy for year-on-year percentage growth. See Usage for input transformations and forecast execution. ForecastData validates the outturn rows when it is created.