Rule Type: Regression

The Machine Learning function provides rule type Regression to train and use a regression model for prediction. The idea of a regression model is to define the relationship between input data and target field using training data and the specific functional form learned depends on the choice of model.

The goal of a regression model is to learn to predict an output based on an input set of features. Taking advantage of gradient boosting technique used by SAP HANA automated predictive library (APL), a regression predicts a target field based on influenced fields. Gradient boosting is a machine learning algorithm to find the shortcomings in the previous predictions and combines base learners by sequentially minimizing the difference between the actual and predicted values. It mainly deals with large volumes of data to make a prediction with high prediction power.

Rule Input Fields

You can use the following input fields for configuration:

  • Target Field: Specifies the field that stores target values of the prediction

  • Influence Fields: Specifies the model input fields (features) that are used to find the assumed relationship to the target field

  • Segmented By: Specifies the fields according to which the whole input data is to be segmented. An independent regression model is trained for each segment. This means, an independent segmentation into regression is found. If this list is empty, then the whole input dataset is considered as one segment by default.

  • Order by Fields: Specifies the fields according to that the segmented datasets are to be sorted.

Rule Output Fields

You can use the following output fields for configuration:

  • Predicted Value: Specifies a field that stores the predicted values. It must be a key figure of numeric type.

  • Model: The field returns the unique ID of the trained model for each segment that is being used for predicting. It is a characteristic field type and we recommend you use a length of 30 characters.