Figure 5-2, The Apply Process with Binning

This figure depicts the model apply process in the Java interface when binning is performed. Applying a model to new data is often called "scoring" the data.

Since the data for model build was binned, the data for model apply must also be binned.

The left side of the diagram shows the steps to perform before data can be supplied to the algorithm that will score a new data table. These "pre-processing" steps are termed in the figure "user data preparation." The pre-processing result is the apply data table. The apply data table and the apply output (specified by the user) are shown as boxes with arrows pointing to the algorithm, indicating that they are supplied (as input) to the mining algorithm.

During user data preparation, the user transforms raw data (for example, randomizing, normalizing, etc. ). The user may bin the data manually, which produces (in addition to binned data) bin boundary tables. In cases where there is no clear way to determine optimal or preferred bin boundaries, ODM can perform the binning automatically.

The apply output specification is a set of objects that determine the content of the result table.

The algorithm then applies the model, using the pre-processed data, the apply output settings, the internal bin boundary tables, and the core algorithm.

The result is a table with scores for each data element in the apply data table.

End of description.