Training a Gradient Boosting Model for Salary Prediction

The script initiates a CAS session, loads the 'sashelp.baseball' dataset into CAS memory under the 'casuser' CASLIB, then loads the 'decisionTree' action set. It then uses the 'gbtreeTrain' action to build a Gradient Boosting model. The model is configured with various options such as POISSON distribution, early stopping based on 'LOGLOSS', and variable importance. The trained model is saved as a CAS table 'GRADBOOST3'.

Data Analysis

Type : SASHELP

The source data comes from SAS's built-in 'sashelp.baseball' dataset, which is then loaded and processed in CAS memory under the 'casuser' CASLIB.

1 Code Block

DATA STEP Data

Explanation :
This code block initializes a CAS session and makes all CASLIBs available. A DATA STEP is then used to load the 'sashelp.baseball' dataset into CAS memory under the 'casuser' CASLIB, thereby creating a working copy of the 'baseball' table in CAS memory.

Copied!

1	cas;
2	caslib _all_ assign;
3
4	DATA casuser.baseball;
5	SET sashelp.baseball;
6	RUN;

2 Code Block

PROC CAS

Explanation :
This block uses PROC CAS to load the 'decisionTree' action set. This action set provides the necessary actions for building and training decision tree models, including the Gradient Boosting action that will be used later.

Copied!

1	PROC CAS;
2	LOADACTIONSET 'decisionTree';
3	QUIT;

3 Code Block

PROC CAS Data

Explanation :
This block sets the default CASLIB to 'casuser' and then uses the 'gbtreeTrain' action from the 'decisionTree' action set via PROC CAS. This action trains a Gradient Boosting model on the 'baseball' table (casuser.baseball), targeting the 'logSalary' variable. It specifies a list of numerical and nominal input variables, uses a POISSON distribution, and includes options for early stopping (based on LOGLOSS), name encoding, greedy selection, handling missing values, Lasso regularization, leaf size, learning rate, and variable importance calculation. The trained model is saved in a new CAS table named 'GRADBOOST3'.

Copied!

1	options caslib=casuser;
2
3	PROC CAS;
4	decisionTree.gbtreeTrain /
5	TABLE={name="baseball"}
6	target="logSalary"
7	casOut={name="GRADBOOST3", replace=true}
8	inputs={"nAtBat",
9	"nHits",
10	"nHome",
11	"nRuns",
12	"nRBI",
13	"nBB",
14	"YrMajor",
15	"CrAtBat",
16	"CrHits",
17	"CrHome",
18	"CrRuns",
19	"CrRbi",
20	"CrBB",
21	"nOuts",
22	"nAssts",
23	"nError",
24	"Division",
25	"League",
26	"Position"}
27	nominals={"Division","League","Position"}
28	distribution="POISSON"
29	earlyStop={metric="LOGLOSS"}
30	encodeName=TRUE
31	greedy=TRUE
32	includeMissing=TRUE
33	lasso=1
34	leafSize=5
35	learningRate=.1
36	m=5
37	varImp=TRUE
38	;
39	QUIT;

This material is provided "as is" by We Are Cas. There are no warranties, expressed or implied, as to merchantability or fitness for a particular purpose regarding the materials or code contained herein. We Are Cas is not responsible for errors in this material as it now exists or will exist, nor does We Are Cas provide technical support for it.

Retour à la liste