RECOMMEND Procedure

Overview: RECOMMEND Procedure

Purpose of the RECOMMEND Procedure

PROC RECOMMEND is an interactive procedure that executes the statements within RUN blocks, similar to the IMSTAT procedure. All tasks in a SAS LASR Analytic Server-based recommender system can be performed from the RECOMMEND procedure.

You use the RECOMMEND procedure to perform the following tasks:

start a recommender system in a SAS LASR Analytic Server
connect to an existing recommender system
interact with one or more recommender systems in the same SAS LASR Analytic Server
filter data using a WHERE clause
remove a recommender system from a SAS LASR Analytic Server
populate a recommender system with content-based data for use with the cluster-based method or with rating data for collaborative filtering
add methods to a recommender system
remove methods from a recommender system
define and optimize ensemble methods
hold a subset of the data to use for model validation
obtain recommendations for users

Working with Recommender Systems

Identify a recommender system in the SAS LASR Analytic Server by a two-level name, similar to a LIBNAME.MEMBER construct. The name, which is case-insensitive, must be unique among the recommender systems in the SAS LASR Analytic Server. It is possible for an in-memory table and a recommender system to have the same name, because the table and application name spaces are separate. However, as a best practice, avoid using the same names to prevent confusion for users.

You can specify the recommender system that you want to work with through options in the PROC RECOMMEND statement or through options on other statements. When you specify the recommender system in the PROC RECOMMEND statement, the procedure uses that system as the default during execution. You can explicitly call other recommender systems in statements within the RECOMMEND procedure. If no application name is specified, the default recommender system is RECOM.SYSTEM.

You add tables to a recommender system to identify the user content, item content, and ratings. Typically, you load a user-item-ratings table into the recommender system. The SAS LASR Analytic Server derives a number of internal tables from the tables that you provide. The subsequent operations of the recommender system use the internal tables only. Make sure the SAS LASR Analytic Server has sufficient resources to build the recommender system and its associated tables. You can drop the user-item-ratings table after the recommender system has been set up.

Access to a recommender system is governed by signature files in the same way that those files are used to control access to SAS LASR Analytic Server tables. However, a recommender system contains multiple derived tables. For example, granting Read access to a recommender system implies Read access to all associated tables for that system.