Scheduling the Batch Forecast Training task

Before scheduling the Batch Forecast Training task, you must configure the file location where the forecasting model files are stored.

See File location for training model files.

  1. Select Maintenance > System Administration > Job Scheduler.
  2. Click Add Schedule.
    The Task page is displayed.
  3. Specify a descriptive name in the Task Description field.
  4. At the Task Type section, select Java Task, then select BatchForecastTraining from the drop-down list.
  5. Assign a specific scheduler to run the task in the Task Affinity field.
  6. In the Scheduled Time section, specify this information:
    Run at
    Specify the time to run the task.
    Timezone
    Select the applicable time zone using the lookup. By default, the Timezone field displays your actual local time zone. If you leave the Timezone field as the default, then the Run at time will be run at the employee’s local time.
  7. In the Scheduling Time section, select one of these options:
    Once
    Schedule the task to run only once.
    Daily
    Schedule the task to run by intervals of days and minutes. If using this option, then you must configure one of these options:
    • Specify the number of days between runs in the Every Day(s) field.
    • Specify the number of seconds between runs in the Interval Second(s) field. You can specify a minimum of 5 seconds between runs. Specify -1 to not use the field.
    Weekly
    Schedule the task to run by day of week. If using this option, then select the check boxes to indicate which days of the week the task will be run on.
    Monthly
    Schedule the task to run by month and by day of month. If using this option, then you must configure one of these options:
    • Select the Day Of option and type the day of month to indicate which date the task runs.
    • Select the Of option and select the week of the month and the day of week to indicate which day the task runs. For example, you can select Second and Thursday to run the task every second Thursday of the month.
    • Select the check boxes to indicate which months of the year the task runs.
  8. At the Blackout Period section, select None.
  9. In the Scheduling Range section, specify this information:
    Start On
    Select the date that you want the task run to begin.
    End By
    Specify the end date of the task run. Or, if the task run has an infinite end date, leave the field blank.
  10. Click Submit, and click OK at the confirmation prompt.
    Your task is added to the list of scheduled tasks in the Task Schedules page.
  11. For your Batch Forecast Training task, click Parameters to specify job parameters.
  12. In the Location field, select the location to train using this task. When the task runs, it trains any drivers that have been configured to use machine learning for the selected location and any of its sub-locations.
  13. The options in the Time Range field control how the historical date range is selected for the actual results data that is used to train the model. Select Year Offset to specify the date range as a number of years before the current date or a specified date, or select Date Range to use a specific date range.
  14. If you selected Year Offset, then specify this information:
    Range in Years
    Specify the length of the date range in years for the historical data that is used to train.
    Before Date
    Leave this field blank to use the current date as the end date for the historical data, or specify an end date for the historical data. The historical date range goes back from this date by the number of years specified in the Range of Years field. For example, if the Before Date is May 9, 2022 and the Range in Years is 5, the historical date range is May 9, 2017 to May 9, 2022.
  15. If you selected Date Range, then specify the Start Date and End Date for the historical data date range.
  16. Specify this information:
    Batch Size
    Specify the number of stores for which the task can simultaneously generate models. Increasing the batch size can decrease the total time required to generate models for all the stores being trained, but requires additional CPU resources from the job scheduler server. Training a large number of stores at the same can result in reduced performance on the job scheduler, and application server if it is on the same physical server as the job scheduler. It is recommended to specify a Batch Size of 10 unless you have special requirements.
    Workmail Notification
    Select the users to be notified when the job is complete.
  17. Click Submit to save the task parameters and click OK to confirm.