# ARIMA Forecast

{% hint style="info" %}
This tool is currently in Beta and is still being tested. Want to learn more? Like to provide feedback? Please reach out to <support@cascade.io>
{% endhint %}

*ARIMA*, short for 'AutoRegressive Integrated Moving Average', allows you to forecast a time series using the series past values.

### Input/Output

| Inputs                                                                                                     | Outputs                                                                                   |
| ---------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| **Time Series Table** - Time series dataset that includes a date column and at least one numerical column. | Table with a new column per forecasted time series values based on the time series input. |

### Options

| Option                          | Description                                                                                                                                                                                                       |
| ------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Date/Time Column**            | Column to be used as the date time reference for the forecast.                                                                                                                                                    |
| **Forecast Columns**            | Selection of columns to be used to during training and prediction of the ARIMA forecast model.                                                                                                                    |
| **Select all numerical fields** | Replaces the "Forecast Columns" prompt above (cannot select both). Instead of selecting individual columns in your input table, choosing this option will automatically select all columns with numerical formats |
| **Interval**                    | Time frequency of the date/time column specified.                                                                                                                                                                 |
| **Train-Test Split**            | Defines the split between training and test data.                                                                                                                                                                 |
| **Periods to Forecast**         | Number of periods to forecast, starting from train-test split.                                                                                                                                                    |

### Tuning Options

| Option                  | Description                                                                                                                                             |
| ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Tuning Parameters**   | When set to auto, the ARIMA model will choose parameters automatically. When set to manual, you will have access to specify the ARIMA model parameters. |
| **p**                   | The number of lag observations included in the model, also called the lag order.                                                                        |
| **d**                   | The number of times that the raw observations are differenced, also called the degree of differencing.                                                  |
| **q**                   | The size of the moving average window, also called the order of moving average.                                                                         |
| **P**                   | The number of lag observations included in the seasonal part of the model, also called the lag order.                                                   |
| **D**                   | The number of times that the raw observations are differenced, also called the degree of differencing, for the seasonal part of the model.              |
| **Q**                   | The size of the moving average window, also called the order of moving average, for the seasonal part of the model                                      |
| **s**                   | Seasonal differencing--refers to the number of periods in each season.                                                                                  |
| **Prediction Interval** | Interval for the testing significance. Specify values 0.0 to 1.0.                                                                                       |