Scoring
The purpose of scoringis to valuate data records. Here you can choose from three function types with which to perform the valuation:
- Weighted score tables
-
Regression Analysis
- Linear regression
- Nonlinear regression (implemented with multilinear splines)
The data that you use to train a model can be taken from any other system, provided that the system can extract data into SAP NetWeaver BW. Likewise, you can apply the same valuation to any data that has been extracted into SAP NetWeaver BW.
You can use queries in SAP NetWeaver BW to access the data with which you wish to calculate scores. For the function types Linear Regression and Nonlinear Regression, training requires data where the numeric target value is known.
You can make the following settings in a model for the Scoring method:
You use the model parameters to specify the function type.
You use the model fields to specify the attributes you wish to use in each case. You specify for each model field whether the corresponding attribute takes discrete or continuous values or whether it is a key field. You also specify the target value.
Some more settings depend on the function type:
- In the case of weighted score tables, you also need to specify the range of values for the individual model fields and enter a weighting.
- For both the regression analysis functions, you can enter or automatically determine an interval for continuous model fields, and you can specify for discrete model fields the attribute values to be considered or simply select those occurring most frequently. Furthermore, several options are open to you for dealing with outliers and missing values.
When using historic data to train the function types Linear Regression and Nonlinear Regression, the system determines the effect of the model field values on the numeric target value. When you train the function type Weighted Score Tables, the function is determined directly, independently of historic data. With all three function types, the system applies the calculated function on the data from the training source. You can display the result in graphical format and export the visualization data into an Excel workbook.