The HP4SCORE Procedure
Overview: HP4SCORE Procedure
The HP4SCORE procedure is a high-performance procedure that can either score a data set or estimate the importance of variables by using a forest predictive model that was previously trained by the HPFOREST procedure, which is described in Chapter 7, The HPFOREST Procedure.
The forest predictive model is an ensemble of hundreds of decision trees that are used to predict a target. The target can have either an interval, nominal, or ordinal measurement level. Each decision tree consists of a sequence of rules that are applied to the observation to arrive at the prediction.
The final prediction is either of the following:
For a target that has an interval measurement level, the final prediction is an average of the individual predictions.
For a target that has a nominal or ordinal measurement level, the final prediction is derived from the average of the individual posterior probabilities.
The HP4SCORE procedure is designed to be used after the HPFOREST procedure creates a model and saves it in a binary model file. PROC HP4SCORE applies the binary model file to score or evaluate the importance of the variables in a specified data set. The variables in the input data set for PROC HP4SCORE must have the same attributes as the variables in the input data set that PROC HPFOREST used to create the model. If the attributes do not match, the HP4SCORE procedure stops and returns an error.
The HP4SCORE procedure can score the data in concurrent threads that run in parallel when it executes in single-machine or distributed mode. For information about how to configure the execution mode, see the section Processing Modes in Chapter 2, Shared Concepts and Topics.