> ## Documentation Index
> Fetch the complete documentation index at: https://nixtlaverse.nixtla.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Data

> Utilies for generating time series datasets

### `generate_series`

```python theme={null}
generate_series(n_series, freq='D', min_length=50, max_length=500, n_static_features=0, equal_ends=False, with_trend=False, static_as_categorical=True, n_models=0, level=None, engine='pandas', seed=0, n_hist_exog=0, n_futr_exog=0, h=0)
```

Generate Synthetic Panel Series.

**Parameters:**

| Name | Type | Description | Default |
| - | - | - | - |
| `n_series` | <code>[int](#int)</code> | Number of series for synthetic panel. | *required* |
| `freq` | <code>[str](#str)</code> | Frequency of the data (pandas alias). Seasonalities are implemented for hourly, daily and monthly. Defaults to 'D'. | <code>'D'</code> |
| `min_length` | <code>[int](#int)</code> | Minimum length of synthetic panel's series. Defaults to 50. | <code>50</code> |
| `max_length` | <code>[int](#int)</code> | Maximum length of synthetic panel's series. Defaults to 500. | <code>500</code> |
| `n_static_features` | <code>[int](#int)</code> | Number of static exogenous variables for synthetic panel's series. Defaults to 0. | <code>0</code> |
| `equal_ends` | <code>[bool](#bool)</code> | Series should end in the same timestamp. Defaults to False. | <code>False</code> |
| `with_trend` | <code>[bool](#bool)</code> | Series should have a (positive) trend. Defaults to False. | <code>False</code> |
| `static_as_categorical` | <code>[bool](#bool)</code> | Static features should have a categorical data type. Defaults to True. | <code>True</code> |
| `n_hist_exog` | <code>[int](#int)</code> | Number of historic exogenous variables. Must be non-negative. Creates columns named `hist_exog_{i}` only in the historic dataframe. Defaults to 0. | <code>0</code> |
| `n_futr_exog` | <code>[int](#int)</code> | Number of future exogenous variables. Must be non-negative. Each variable is generated once and split between the historic and future dataframes. Creates columns named `futr_exog_{i}` in both dataframes. Defaults to 0. | <code>0</code> |
| `h` | <code>[int](#int)</code> | Number of future periods to generate per series when `n_futr_exog` is greater than 0. Defaults to 0. | <code>0</code> |
| `n_models` | <code>[int](#int)</code> | Number of models predictions to simulate. Defaults to 0. | <code>0</code> |
| `level` | <code>list of float</code> | Confidence level for intervals to simulate for each model. Defaults to None. | <code>None</code> |
| `engine` | <code>[str](#str)</code> | Output Dataframe type. Defaults to 'pandas'. | <code>'pandas'</code> |
| `seed` | <code>[int](#int)</code> | Random seed used for generating the data. Defaults to 0. | <code>0</code> |

**Returns:**

| Type | Description |
| - | - |
| <code>[Union](#typing.Union)\[[DataFrame](#utilsforecast.compat.DataFrame), [Tuple](#typing.Tuple)\[[DataFrame](#utilsforecast.compat.DataFrame), [DataFrame](#utilsforecast.compat.DataFrame)]]</code> | pandas or polars DataFrame, or tuple of DataFrames: Synthetic panel with columns \[`unique_id`, `ds`, `y`] and exogenous features. When `n_futr_exog` is greater than 0, returns `(df, futr_df)`, where `futr_df` has `h` future timestamps per series. |


## Related topics

- [Intermittent Data](/neuralforecast/docs/tutorials/intermittent_data.html.md)
- [Example Data](/neuralforecast/utils.html.md)
- [Data Requirements](/neuralforecast/docs/getting-started/datarequirements.html.md)
- [Intermittent or Sparse Data](/statsforecast/docs/tutorials/intermittentdata.html.md)
- [Multivariate missing data](/synforecast/docs/capabilities/multivariate_missingness.html.md)