Column Mappings (Lift Chart)
You can use the Column Mapping tab of the Mining Accuracy Chart tab of Data Mining Designer to provide information to Analysis Services to describe the data and the mappings that you want to use to test the accuracy of mining models.
The following table lists the tasks that you can complete in the Column Mapping tab, and directs you to other topics that describe how to compete these tasks.
Map columns in the mining structure to columns in the input dataset
Filter the input data
Choose predictable columns to use to create the lift chart
The top half of the Column Mappings tab contains two tables, Mining Structure and Select Input Table(s). Use these tables to select the mining structure that contains the models that you want to validate, and to select the input table that the models are to be validated against. You can select any mining structure in a project, and you can select an input table from any data source view in the project. When you select an input table, the columns in the tables Mining Structure and Select Input Table(s) are automatically mapped together. You can modify the mappings as needed by clicking a column in the Mining Structure table and dragging it to the Select Input Table(s) table. If the input data contains a nested table, you can also include this table by using the Select nested table link. Not all columns have to be mapped together, but the predictable column must always be mapped. Columns that are not mapped are fed as NULL values to the mining model
You can filter the data in the input table by using the filtering table. This means that you create a WHERE clause in the query that is used to collect the data from the input table. The filtering table contains the following columns:
Determines the source of the new column. Possible sources include the mining model, the input tables, a prediction function, or a customized expression.
Determines the specific column or function that is associated with the selection in the Source column.
Works with the And/Or column to group expressions together using parentheses. For example, (expr1 or expr2) and expr3.
Creates logic in the query. For example, (expr1 or expr2) and expr3.
Specifies a condition or user expression that applies to the column. Literals in the condition require single quotation marks.
The Synchronize Prediction Columns and Values option coordinates the predictable attributes in the grid so that, even if they have a different name, they are derived from the same predictable column during model training. If you clear the Synchronize Prediction Columns and Values check box, you can select any valid predictable column and value, and the results are plotted together, even if the results do not make sense.
A mining structure can contain several mining models. The bottom section of the Column Mappings tab contains a table that you can use to select the model to include in the lift chart, the predictable column to chart against, and the state of the predictable column. If you leave the state of the predictable column blank, the lift chart predicts how well the model performs regardless of the state of the predictable column. For more information about the differences between creating lift charts with or without a specified state of the predictable column, see Lift Chart.