Panel Data¶

PanelForecaster fits a model independently to each group in a pandas DataFrame.

Usage¶

import pandas as pd
from durbyn import AutoARIMA, PanelForecaster

# Panel data with multiple groups
data = pd.DataFrame({
    "store": ["A"] * 48 + ["B"] * 48 + ["C"] * 48,
    "date": list(range(48)) * 3,
    "sales": [...],  # sales data for each store
})

# Create a panel forecaster
pf = PanelForecaster(model=AutoARIMA(), groupby="store")

# Fit to each group
pf.fit(data, y_col="sales", m=12, date_col="date")

# Forecast all groups
result = pf.forecast(h=12, level=[80, 95])

# Long-format DataFrame
df = result.to_dataframe()
print(df)

PanelForecaster¶

Constructor¶

Parameter	Type	Description
`model`	`BaseForecaster`	Template model (deep-copied per group)
`groupby`	`str \\| list[str]`	Column(s) to group by

`.fit(data, y_col="y", m=1, date_col=None)`¶

Parameter	Type	Default	Description
`data`	`pd.DataFrame`	required	Panel data
`y_col`	`str`	`"y"`	Target column name
`m`	`int`	`1`	Seasonal period
`date_col`	`str \\| None`	`None`	Date column for sorting within groups

Returns self for method chaining.

`.forecast(h=10, level=None)`¶

Parameter	Type	Default	Description
`h`	`int`	`10`	Forecast horizon
`level`	`list[int] \\| None`	`None`	Confidence levels

Returns a PanelForecastResult.

PanelForecastResult¶

Attributes¶

Attribute	Type	Description
`groups`	`dict[Any, ForecastResult]`	Forecast result per group key

`.to_dataframe()`¶

Returns a long-format pandas.DataFrame with a group column and all forecast columns (step, mean, lower/upper intervals).

Multi-Group Keys¶

When grouping by multiple columns, group keys are tuples:

pf = PanelForecaster(model=AutoARIMA(), groupby=["region", "product"])
pf.fit(data, y_col="sales", m=12)
result = pf.forecast(h=6)

# Access individual group
fc = result.groups[("East", "Widget")]

Error Handling¶

If a model fails to fit or forecast for a group, a warning is issued and that group is skipped.

import warnings

with warnings.catch_warnings(record=True):
    pf.fit(data, y_col="sales", m=12)