Before you Begin

This 30 minute tutorial shows you how to make predictions with Machine Learning (ML) models in EPM Cloud. Before stepping through this tutorial, please review the Importing ML Models hands-on tutorial. The sections build on each other and should be completed sequentially.

Background

With Bring Your Own ML, EPM administrators can import a fully trained ML model and deploy it to a Planning application. Planners can then leverage robust, ML-based forecasting that uses advanced predictive modeling techniques to generate more accurate forecasts.

Data scientists build and train the ML model in a data science tool such as any third party tool or Oracle Data Science Cloud, and save it as a PMML file. Next, EPM administrators incorporate the model into Planning helping planners gain the business value from the trained model:

  1. Administrators import the ML model in PMML format to a Planning application, then define how the input variables and target variable maps to dimension members or cell values in the Planning application.
    Model Mapping

    This step generates automatic Groovy rules that connect the ML model to the Planning application. Two Groovy rules are generated for each ML model definition: one rule to associate with a form or dashboard, which allows users to make predictions on demand, and another to generate large scale predictions in a scheduled job for bulk processing.


    Groovy Rules
  2. Administrators deploy the ML model in a Planning application by associating the Groovy rule to relevant action menus, forms, or dashboards. Administrators can also create a job to run the Groovy rule in a batch process.
  3. Planners leverage ML-powered business rules in forms to generate predicted values, which are saved on the form. Planners can perform what-if analysis using the generated predictions, or modify predicted values on the form. Planners add value with their expertise and judgement, and then finalize the forecast.
  4. This is an iterative process. As planners make predictions based on the ML model, administrators can measure the performance of the model, and can work with data scientists to update or replace the ML model when needed. Then, administrators re-import and deploy the retrained ML model. When you re-import the retrained ML model, the Groovy rules are regenerated.

In this tutorial, you take the Groovy rule created from an imported ML Model, and add it to a form. Then you make predictions with the ML model in Planning.

What Do You Need?

An EPM Cloud Service instance allows you to deploy and use one of the supported business processes. To deploy another business process, you must request another EPM Enterprise Cloud Service instance or remove the current business process.

You must have:

  • Service Administrator access to an EPM Enterprise Cloud Service instance.
  • Upload and import this snapshot into your instance.

Activating Navigation Flows

In this section, you activate the ML Tutorial navigation flow so you can work with ML Models.

  1. On the home page, click Tools, then Navigation Flows.
    home page
  2. For the ML Tutorial row, under Active, click Inactive.
    Navigation Flow page

    The ML Tutorial navigation flow is activated, and the Default flow is inactivated.


    Navigation Flow page
  3. Click Home (Home).
  4. Click Navigation Flow (Navigation Flow), and click ML Tutorial.
    Navigation Flow

    The ML Tutorial Navigation Flow is displayed.


    home page

Assigning Values to User Variables

User Variables were added when the business process was created. User variables act as filters in forms, enabling planners to focus only on certain members. In this section, you set values for user variables.

  1. On the home page, click Navigator (Navigator), and under Tools, click User Preferences.
    Navigator menu
  2. Under Preferences, click User Variables.
    Selecting User Variable Values
  3. For each variable, click its Member Selector (Member Selector) to select a member as the variable's value:
    User Variable Member
    Infolet Entity NA
    Currency USD
    Entity Sales US - West
    Reporting Currency USD
    Scenario Actual
    Version Working
    Years FY22
    Expense Account Total Marketing Expense
    Expense Drivers Marketing Expense Drivers
    Product Line 250-Servers
  4. Verify your selections and click Save.
    User variables with selected Members
  5. At the information message, click OK.
    Information message
  6. Return to the home page. Click Home (Home).

Reviewing Data for Key Drivers

In this section, you review key drivers for selling price, promotion spend, and industry volume.

  1. On the home page, click IPM then Volume Forecasting.
    home page

    2. The Product Volume Forecasting dashboard is displayed. On the top form, product predictions are displayed. This includes volume by product category for all products. We plan to predict product volume for the first six months of FY22. On the bottom form, you can view product drivers.


    Product Volume Forecasting
  2. On the bottom form for Drivers, from the drop-down list, select Average Selling Price and on the right side of the form, click go (Go).
    Product Volume Forecasting dashboard

    The average selling price is one of the drivers used to predict volume.

  3. On the bottom form for Drivers, from the drop-down list, select Advertising and Promotion and on the right side of the form, click go (Go).
    Product Volume Forecasting dashboard

    Advertising and Promotion data is used to predict volume.

  4. On the bottom form for Drivers, from the drop-down list, select Industry Volume and on the right side of the form, click go (Go).
    Product Volume Forecasting dashboard

    Industry Volume data is used to predict volume.

Deploying an ML Model to Planning

In this section, you deploy the ML Model to Planning by creating an action menu and selecting the Groovy rule to run from that action menu. You then add the action menu to a form. You can also run an ML Groovy rule as a batch job.

Creating an Action Menu

  1. Click Navigator (Navigator) and under Create and Manage, click Action Menus.
    Navigator Menu
  2. Click Create Menu (Create Menu).
  3. Form Menu Name, enter Predict Volume with ML, and click OK.
    Create Menu
  4. Ensure that Predict Volume with ML is selected, and click Edit (Edit).
  5. Click Add Child (Add Child).
    Edit Action Menu
  6. Enter or select the following information, and click Save.
    Action Menu Details
  7. At the Information message, click OK.
    Information Message

Adding an Action Menu to a Form

  1. Click Navigator (Navigator) and under Create and Manage, click Forms.
    Navigator Menu
  2. Under Folders, expand Library, and Financials and then select Financials - Planbook - Revenue.
    Forms and Folders
  3. In the Content panel on the right, select PMML Forecast V2 Inputs, and click Edit (Edit).
    Form and Ad Hoc Grid Management
  4. Click the Other Options tab.
    Properties Tab
  5. In Context Menus, for Available Menus, locate and select Predict Volume with ML, and click Edit (Move Selected Items to Selected Menus).
    Other Options
  6. To save and close the form, click Finish.
    Other Options
  7. Close Form and Ad Hoc Grid Management by clicking X in the top right corner.
    Form and Ad Hoc Grid Management

Running Predictions

In this section you predict volume for the next six months based on key drivers. This approach differs from classic predictive planning which typically uses one measure as a basis for the prediction. Instead, Machine Learning uses a multi-variate approach where you predict product volume by using multiple contributing input drivers. On the Product Volume Forecasting dashboard, you use the drivers on the bottom form to predict product volume on the top form. Product volume predictions are made based on industry volume, average selling price and advertising and promotion.


Product Volume Forecasting

  1. Click Navigator (Navigator) and under IPM, click Volume Forecasting.
    Navigator
  2. In the top form, scroll right to see January through June of FY22.
    Product Volume Forecasting

    The ML Model will be used to predict the first six months of FY22.


    Product Volume Forecasting

  3. On the Drivers form, right-click a cell such as at the intersection of Smart Phone 4 in and Jun FY21, and select Predict Volume with ML.
    Predict Volume with ML

    This triggers the prediction rule that in turn runs the rule using the ML model that was imported and deployed in the planning application.

    The rule picks the input drivers and gets the output which is populated back into the planning form. Planners can run through various what-if scenarios to adjust their forecasts and plans, or adjust the predicted values.

    Note:

    The rule runs in the context of the form, predicting values only for cells on the form. Security is honored so that planners see predictions only for intersections to which they have access.

    The predictions are displayed in the top form.

    Product Volume Forecasting

    Note:

    As planners make predictions based on the ML model, administrators can measure the performance of the model, and can work with data scientists to update or replace the ML model when needed. Then, administrators re-import and deploy the retrained ML model. When you re-import the retrained ML model, the Groovy rules are regenerated.

Learn More