Skip to content

Forecast

The “Forecast” extension offers the forecasting capability as part of Predict offering, which offers an out-of-the-box model building to forecast your future time series data values, using advanced artificial intelligence and machine learning technologies.

It uses time series data from asset as dataset and does easy configuration for forecast model building and execution. It also visualizes forecasting results which could be further utilized for your targeted use cases like comparing with threshold value, reviewing historic forecasting records for model selection.

This extension also allows you to predict the future data on multiple instances and compare them for analysis. This comparison of historic forecasting records assists in the better model selection.

User interface

The following figure shows the "Forecast" screen:

UI

① Dataset information
② Select the target variable for creating the forecast
③ Select predictor variables and timeseries type
④ Select the Configuration parameters

Generating the Forecast model

To generate the forecast report, proceed with the following steps:

  1. Open "Predict 2.0" in the "Analyze" section.
  2. Select the required "Use Case" and click "Forecast Jobs".
  3. Select the targeted variable of your choice.
    The forecast prediction is prepared for the selected target variable.
  4. Select the input variables of your choice.
  5. Configure the below settings to generate the forecast report.
    • Prediction Horizon: Number of samples or time period in the future for forecasting
    • Validation Length: Back-testing length from dataset for accuracy metrics generation
    • Confidence Interval: A measurement of percentage to represent the confidence level
    • Allow offset: A boolean value that decides whether to use offsets of predictors in model
  6. Click "Generate Forecast".

The generated forecast model can be used to predict the future data for the selected variables of the asset.

Forecast report

Scheduling the forecast model for production

It is possible to plan and schedule the production of the forecast model which can be executed and evaluated on the frequent intervals as per the preferred time range.

To schedule the production run of the forecast model, proceed with the following steps:

  1. Open "Predict 2.0" in the "Analyze" tab.
  2. Select the "Use Case", and click "Forecast".
  3. Select the preferred model from the list of available models.
  4. Click "Schedule for production".
  5. Select the preferred frequency for executing the models.
    The forecast models can be executed as per the below intervals:
    • Minutes
    • Hourly
    • Daily
    • Weekly
  6. Select the preferred date range.
  7. Click "Save".

    Schedule for production

Once the forecast is scheduled for production, the forecast reports are generated as per the schedule. The generated forecast report can be used to predict the future data for the selected variables of the asset. To view the forecast report, click "Production forecasts" and select the preferred report from the "detail result for forecast execution" drop-down.

Forecast report

Saving the Forecast report in IoT store

Once the forecast reports are generated as per the schedule, then the aggregates (mean, min and max) are calculated based on the forecast prediction results. This report containing aggregated data is uploaded back into Insights Hub IoT store. Rules engine can be configured to generate events based on Forecast prediction results.

Comparing the Forecast report

Once the forecast report is generated, you can make the required changes in the "Forecast Configurations" and click "Regenerate Forecast" and compare the two reports.

To compare the forecasting reports, select the required forecasting reports and click "Compare".

Compare

The following figure shows the comparison of the two forecast reports:

Two forecast report

Deleting the Forecast Reports

To delete the forecast reports, select the required forecast reports from the list, click threedotsicon and select "Delete".


Last update: September 23, 2024