Using Scoring Functions to Run Scoring Models

How to Run a Scoring Model Using Scoring Functions

The %INDB2_PUBLISH_MODEL macro creates the files that are needed to build the scoring functions and publishes the scoring functions with those files to a specified database in DB2. Only the EM_ output variables are published as DB2 scoring functions. For more information about the EM_ output variables, see Fixed Variable Names.
Note: Secure File Transfer Protocol (SFTP) is used to transfer the source files to the DB2 server during the publishing process. Certain software products that support SSH-2 or SFTP protocols must be installed before you can use the publishing macros. For more information, see Configuring SSH Client Software in UNIX and Windows Environments for Use with the SFTP Access Method in SAS 9.2, SAS 9.3, and SAS 9.4 located at http://support.sas.com/techsup/technote/ts800.pdf.
To run the scoring model using scoring functions, follow these steps.
  1. Run the %INDB2_PUBLISH_MODEL macro.
    The %INDB2_PUBLISH_MODEL macro uses some of the files that are created by the SAS Enterprise Miner Score Code Export node: the scoring model program (score.sas file), the properties file (score.xml file), and (if the training data includes SAS user-defined formats) a format catalog.
    The %INDB2_PUBLISH_MODEL macro performs the following tasks:
    • takes the score.sas and score.xml files and produces the set of .c and .h files. These .c and .h files are necessary to build separate scoring functions for each of a fixed set of quantities that can be computed by the scoring model code.
    • if a format catalog is available, processes the format catalog and creates an .h file with C structures. These files are also necessary to build the scoring functions.
    • produces a script of the DB2 commands that are used to register the scoring functions on the DB2 database.
    • transfers the .c and .h files to DB2 using SFTP.
    • calls the SAS_COMPILEUDF function to compile the source files into object files, links to the SAS formats library, and copies the new object files to db2path/sqllib/function/SAS, where db2path is the path that was defined during installation. The object filename is dbname_schemaname_modelname_segnum, where segnum is a sequence number that increments each time the model is replaced or re-created. The object file is renamed to avoid library caching in DB2.
    • calls the SAS_DELETEUDF function to remove existing object files.
    • uses the SAS/ACCESS Interface to DB2 to run the script to create the scoring functions with the object files.
    The scoring functions are registered in DB2 with shared object files, which are loaded at run time. These functions are stored in a permanent location. The SAS object files and the SAS formats library are stored in the db2path/sqllib/function/SAS directory, where db2path is the path that was defined during installation. This directory is accessible to all database partitions.
    DB2 caches the object files after they are loaded. Each time that the updated objects are used, one of the following actions must occur:
    • The database must be stopped and restarted to clean up the cache.
    • The object files need to be renamed and the functions reregistered with the new object filenames.
    The SAS publishing process automatically handles the renaming to avoid stopping and restarting the database.
    Note: You can publish scoring model files with the same model name in multiple databases and schemas. Because object files for the SAS scoring function are stored in the db2path/sqllib/function/SAS directory, the publishing macros use the database, schema, and model name as the object filename to avoid potential naming conflicts.
  2. Use the scoring functions in any SQL query.

Scoring Function Names

The names of the scoring functions that are built in DB2 have the following format:
modelname_EM_outputvarname
modelname is the name that was specified in the MODELNAME argument of the %INDB2_PUBLISH_MODEL macro. modelname is always followed by _EM_ in the scoring function name. For more information about the MODELNAME argument, see %INDB2_PUBLISH_MODEL Macro Syntax.
outputvarname is derived from the names of the EM_ output variables in the score.xml file that is generated from the SAS Enterprise Miner Score Code Export node. For more information about the score.xml file, see Fixed Variable Names.
One scoring function is created for each EM_ output variable in the score.xml file. For example, if the scoring model DATA step program takes ten inputs and creates three new variables, then three scoring functions are defined. Each scoring function has the name of an output variable. For example, if you set MODELNAME=credit in the %INDB2_PUBLISH_MODEL macro and the EM_ output variables are “EM_PREDICTION”, “EM_PROBABILITY”, and “EM_DECISION”, then the name of the scoring functions that are created would be “credit_EM_PREDICTION”, “credit_EM_PROBABILITY”, and “credit_EM_DECISION”.
Note: A scoring function name cannot exceed 128 characters.
CAUTION:
When the scoring function is generated, the names are case insensitive.
Consequently, if you have model names “Model01” and “model01”, and you create two scoring functions, the second scoring function overwrites the first scoring function.

Viewing the Scoring Functions

The scoring functions are available to use in any SQL expression in the same way that DB2 built-in functions are used. For an example, see Using Scoring Functions to Run a Scoring Model.
There are four ways to see the scoring functions that are created:
  • From DB2, log on to the database using the DB2 client tool (command line processor) and submit an SQL statement. The following example assumes that the model name that you used to create the scoring functions is mymodel and the DB2 installation instance is located in /users/db2v9. The first line of code executes a db2profile script. The script sets the DB2 environment variables so that the DB2 command line processor (CLP) can execute.
    >./users/db2v9/sqllib/db2profile
    >db2
    db2 => connect to database user username using password
    db2 => select * from syscat.functions where funcname like '%MYMODEL%'
  • From SAS, use SQL procedure code that produces output in the LST file. The following example assumes that the model name that you used to create the scoring functions is mymodel.
    proc sql noerrorstop;
      connect to db2 (user=username pw=password db=database);
      
    select *
       from connection to db2
          (select * from syscat.functions where funcname like '%MYMODEL%');
       disconnect from db2;
    quit;
    
  • Look at the SampleSQL.txt file that is produced when the %INDB2_PUBLISH_MODEL macro is successfully run. This file can be found in the output directory (OUTDIR argument) that you specify in the macro.
    The SampleSQL.txt file contains basic SQL code that can be used to run your score code inside DB2. Please note that you must modify the sample code before using it. Otherwise, the sample code returns an error.
    For example, the SampleSQL.txt file refers to an ID column in allmush1_intab that is populated with a unique integer from 1 to n. n is the number of rows in the table.
    Note: The ID column uniquely identifies each row. You would replace the ID column with your own primary key column.
    Note: The function and table names must be fully qualified if the function and table are not in the same schema.
    The following example assumes that the model name that you used to create the scoring functions is allmush1.
    drop table allmush1_outtab;
    create table allmush1_outtab(
     id integer
    ,"EM_CLASSIFICATION" varchar(33)
    ,"EM_EVENTPROBABILITY" float
    ,"EM_PROBABILITY" float
    );
    insert into allmush1_outtab(
     id
    ,"EM_CLASSIFICATION"
    ,"EM_EVENTPROBABILITY"
    ,"EM_PROBABILITY"
    )
    select id,
     allmush1_em_classification("BRUISES"
    ,"CAPCOLOR"
    ,"GILLCOLO"
    ,"GILLSIZE"
    ,"HABITAT"
    ,"ODOR"
    ,"POPULAT"
    ,"RINGNUMB"
    ,"RINGTYPE"
    ,"SPOREPC"
    ,"STALKCBR"
    ,"STALKROO"
    ,"STALKSAR"
    ,"STALKSHA"
    ,"VEILCOLO")
      as "EM_CLASSIFICATION",
     allmush1_em_eventprobability("BRUISES"
    ,"CAPCOLOR"
    ,"GILLCOLO"
    ,"GILLSIZE"
    ,"HABITAT"
    ,"ODOR"
    ,"POPULAT"
    ,"RINGNUMB"
    ,"RINGTYPE"
    ,"SPOREPC"
    ,"STALKCBR"
    ,"STALKROO"
    ,"STALKSAR"
    ,"STALKSHA"
    ,"VEILCOLO")
      as "EM_EVENTPROBABILITY",
     allmush1_em_probability("BRUISES"
    ,"CAPCOLOR"
    ,"GILLCOLO"
    ,"GILLSIZE"
    ,"HABITAT"
    ,"ODOR"
    ,"POPULAT"
    ,"RINGNUMB"
    ,"RINGTYPE"
    ,"SPOREPC"
    ,"STALKCBR"
    ,"STALKROO"
    ,"STALKSAR"
    ,"STALKSHA"
    ,"VEILCOLO")
      as "EM_PROBABILITY"
    from allmush1_intab ;
    
  • You can look at the SAS log that is created when the %INDB2_PUBLISH_MODEL macro was run. A message that indicates whether a scoring function is successfully or not successfully executed is printed to the SAS log.

Using Scoring Functions to Run a Scoring Model

The scoring functions are available to use in any SQL expression in the same way that DB2 built-in functions are used.
The following example code creates the scoring functions.
%let indconn = server=db2base user=user1 password=open1 database=mydb;
%indb2_publish_model( dir=C:\SASIN\baseball1, modelname=baseball1);
The %INDB2_PUBLISH_MODEL macro produces a text file of DB2 CREATE FUNCTION commands as shown in the following example.
Note: This example file is shown for illustrative purposes. The text file that is created by the %INDB2_PUBLISH_MODEL macro cannot be viewed and is deleted after the macro is complete.
CREATE FUNCTION baseball1_EM_eventprobablility
(
"CR_ATBAT" float,
"CR_BB" float,
"CR_HITS" float,
"CR_HOME" float,
"CR_RBI" float,
"CR_RUNS" float,
"DIVISION" varchar(31),
"LEAGUE" varchar(31),
"NO_ASSTS" float,
"NO_ATBAT" float,
"NO_BB" float,
"NO_ERROR" float,
"NO_HITS" float,
"NO_HOME" float,
"NO_OUTS" float,
"NO_RBI" float,
"NO_RUNS" float,
"YR_MAJOR" float
)
RETURNS varchar(33)
LANGUAGE C
NO SQL
PARAMETER STYLE SQL
DETERMINISTIC
FENCED THREADSAFE
NO EXTERNAL ACTION
ALLOW PARALLEL
NULL CALL
EXTERNAL NAME '/users/db2v9/sqllib/function/SAS/
   dbname_username_baseball1.so!baseball1_em_ eventprobablility'
After the scoring functions are installed, they can be invoked in DB2 using SQL, as illustrated in the following example. Each output value is created as a separate function call in the select list.
select baseball1_EM_eventprobability
( 
"CR_ATBAT",
"CR_BB",
"CR_HITS",
"CR_HOME",
"CR_RBI",
"CR_RUNS",
"DIVISION",
"LEAGUE",
"NO_ASSTS",
"NO_ATBAT",
"NO_BB",
"NO_ERROR",
"NO_HITS",
"NO_HOME",
"NO_OUTS"
) as homeRunProb from MLBDB2;