Importing an ML Model
Import a fully trained ML model into a Planning application to prepare it for use by business users.
Prerequisite: Before you can import the ML model, the data science team must build, train, and save the ML model as a PMML file.
To import an ML model to a Planning application:
- From the Home page, click IPM and then click ML Models.
- Click Import, and then drag and drop the PMML file, or browse
to it and select it.
On the Import Model page, you see information about the PMML file, such as the target column (the variable to be predicted using the ML model) and the training date.
- Enter a model name and description, and then click Next.
- On the Generate Rule page, enter information that will
generate a Groovy rule to associate with forms or dashboards:
In Model Mapping, select the cube where the ML model will be used and define the scope of data in which to use the ML model by selecting a member or set of members from each dimension.
- Map Input and Output to the
appropriate dimensions and members in the cube and then click
Next.
The Input and Output sections contain the list of input features (features/columns that are used to make predictions) and target feature (column that is expected to be predicted). Planning analyzes the PMML file to generate the list of inputs and outputs.
Input features are independent variables, similar to drivers, that act as input to your system. When you make predictions, the model uses input features to predict your output. In this step, you map the input from the ML model to the output in the Planning cube. Input describes how to extract the data from the ML model. Output defines the target measure you want to predict and where to paste the predicted values in the Planning application.
For example, product, price, and industry volume, the input features, might be used to predict volume, the output.
-
In the Input area, for each input feature, select an Input Type and if you select Cell Value or Member, select the members or dimensions in the Planning application to map to. Input types:
- Prompt: If you don't have a member or dimension in Planning that maps to this input value from the ML model, when the predicion is made, prompt the user to enter an estimate for the value.
- Cell Value: Map an input feature to one or more dimension members in the Planning cube. For example, the input feature called Price maps to an account member called Price in the Planning application.
- Member: Map an input feature to a dimension in the Planning cube. For example, the Input feature called Product maps to the Product dimension in the Planning application.
-
In the Output area, select an Input Type and if you select Cell Value or Member, select the members or dimensions in the Planning application to map to to store prediction results.
-
- In Analyze Model, review the ML model and then click
Next.
This step represents MLX (Machine Learning Explainability), and extracts additional information about the ML model. For example, review Regression Coefficients to see how the relative impact of key input features is used to predict the output. The height of the bar represents the incremental effect of one unit increase in an input feature on the target variable.
- In Test Model, test the ML model by generating a prediction for a set of sample values. For each Input, enter a sample input value and then click Predict.
- Review the predicted Output value, and then click Save and Close.
- Click Yes to confirm the creation of Groovy Rules.
Two Groovy rules are generated for each ML model definition:
ML_MLModelName_Form
: Use this rule to associate with a form or dashboard, which allows users to make predictions on demand.ML_MLModelName
: Use this rule to generate large scale predictions in a scheduled job for bulk processing.
You can review the generated rules in Calculation Manager. The Groovy rules define the name and location of the PMML file, along with input and output based on the mapping you defined. For more information on using these generated Groovy rules, see Deploying an ML Model to Planners.
Tutorials
Tutorials provide instructions with sequenced videos and documentation to help you learn a topic.
Your Goal | Learn How |
---|---|
Learn how to 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. | ![]() |