/****************************************************************************/
/* This is the first in a series of examples provided to demonstrate the */
/* use of SAS Viya Data Mining and Machine Learning procedures to compose */
/* a program that follows a standard machine learning process of */
/* - loading data, */
/* - preparing the data, */
/* - building models, and */
/* - assessing and comparing those models */
/* */
/* The programs are written to execute in the CAS in-memory distributed */
/* computing engine in the SAS Viya environment. */
/* */
/* This first example showcases how to load local data into CAS */
/****************************************************************************/
/* Define a CAS engine libref for CAS in-memory data tables */
libname mycaslib cas caslib=casuser;
/****************************************************************************/
/* Load data into CAS */
/* */
/* The data set used for this workflow is from a financial services company */
/* that offers a home equity line of credit. The company has extended */
/* several thousand lines of credit in the past, and many of these accepted */
/* applicants have defaulted on their loans. Using demographic and */
/* financial variables, the company wants to build a model to predict */
/* whether an applicant will default. */
/* */
/* The target variable "BAD" indicates whether an applicant defaulted */
/* on the home equity line of credit. */
/* */
/* For execution in the CAS engine, data must be loaded from the local */
/* data set to a CAS table. This code first checks to see if the specified */
/* CAS table exists and then loads data from local data sets in 2 */
/* different ways. After executing this code, you will notice a new */
/* "MYCASLIB" library reference under "Libraries" in the navigation panel */
/* on the left side (note the special icon indicating it is a caslib). */
/* */
/****************************************************************************/
%if not %sysfunc(exist(mycaslib.hmeq)) %then %do;
/* You can load data using a "load" statement in PROC CASUTIL */
proc casutil;
load data=sampsio.hmeq casout="hmeq" outcaslib=casuser;
run;
%end;
%if not %sysfunc(exist(mycaslib.hmeq)) %then %do;
/* You can also load data using a data step */
data mycaslib.hmeq;
set sampsio.hmeq;
run;
%end;1 %studio_hide_wrapper;
77
78 /****************************************************************************/
79 /* This is the first in a series of examples provided to demonstrate the */
80 /* use of SAS Viya Data Mining and Machine Learning procedures to compose */
81 /* a program that follows a standard machine learning process of */
82 /* - loading data, */
83 /* - preparing the data, */
84 /* - building models, and */
85 /* - assessing and comparing those models */
86 /* */
87 /* The programs are written to execute in the CAS in-memory distributed */
88 /* computing engine in the SAS Viya environment. */
89 /* */
90 /* This first example showcases how to load local data into CAS */
91 /****************************************************************************/
92
93 /* Define a CAS engine libref for CAS in-memory data tables */
94 libname mycaslib cas caslib=casuser;
NOTE: Libref MYCASLIB was successfully assigned as follows:
Engine: CAS
Physical Name: c871747f-5b0b-2e45-af8f-e24f0fd568ce
95
96 /****************************************************************************/
97 /* Load data into CAS */
98 /* */
99 /* The data set used for this workflow is from a financial services company */
100 /* that offers a home equity line of credit. The company has extended */
101 /* several thousand lines of credit in the past, and many of these accepted */
102 /* applicants have defaulted on their loans. Using demographic and */
103 /* financial variables, the company wants to build a model to predict */
104 /* whether an applicant will default. */
105 /* */
106 /* The target variable "BAD" indicates whether an applicant defaulted */
107 /* on the home equity line of credit. */
108 /* */
109 /* For execution in the CAS engine, data must be loaded from the local */
110 /* data set to a CAS table. This code first checks to see if the specified */
111 /* CAS table exists and then loads data from local data sets in 2 */
112 /* different ways. After executing this code, you will notice a new */
113 /* "MYCASLIB" library reference under "Libraries" in the navigation panel */
114 /* on the left side (note the special icon indicating it is a caslib). */
115 /* */
116 /****************************************************************************/
117 %if not %sysfunc(exist(mycaslib.hmeq)) %then %do;
118
119 /* You can load data using a "load" statement in PROC CASUTIL */
120 proc casutil;
NOTE: The UUID 'c871747f-5b0b-2e45-af8f-e24f0fd568ce' is connected using session CASAUTO.
121 load data=sampsio.hmeq casout="hmeq" outcaslib=casuser;
NOTE: SAMPSIO.HMEQ was successfully added to the "CASUSER(1440420407@qq.com)" caslib as "HMEQ".
122 run;
123
124 %end;
125
126 %if not %sysfunc(exist(mycaslib.hmeq)) %then %do;
127
128 /* You can also load data using a data step */
129 data mycaslib.hmeq;
130 set sampsio.hmeq;
131 run;
132
133 %end;
134
135
136
137 %studio_hide_wrapper;
145
146
/************************************************************************/
/* This example illustrates various tools for assaying, assessing, */
/* modifying and preparing data prior to modeling. It uses HMEQ */
/* dataset as input and produces HMEQ_PREPPED dataset. The HMEQ_PREPPED */
/* dataset is used in subsequent examples. */
/* */
/* The steps include: */
/* */
/* (1) PREPARE AND EXPLORE */
/* a) Load data set into CAS */
/* b) Explore */
/* c) Impute */
/* d) Identify variables that explain variance */
/* e) Perform a cluster analysis to identify homogeneous */
/* groups in the data */
/* f) Perform principal components analysis to assess collineary */
/* among candidate, interval valued inputs */
/************************************************************************/
/************************************************************************/
/* Setup and initialize for later use in the program */
/************************************************************************/
/* Define a CAS engine libref for CAS in-memory data tables */
libname mycaslib cas caslib=casuser;
/* Specify the data set names */
%let sasdata = sampsio.hmeq;
%let casdata = mycaslib.hmeq;
/* Specify the data set inputs and target */
%let class_inputs = reason job;
%let interval_inputs = clage clno debtinc loan mortdue value yoj derog delinq ninq;
%let target = bad;
%let im_class_inputs = reason job;
%let im_interval_inputs = im_clage clno im_debtinc loan mortdue value im_yoj im_ninq derog im_delinq;
%let cluster_inputs = im_clage im_debtinc value;
/* Specify a folder path to write the temporary output files */
%let outdir = &_SASWORKINGDIR;
/************************************************************************/
/* Load data into CAS if needed. Data should have been loaded in */
/* step 1, it will be loaded here after checking if it exists in CAS */
/************************************************************************/
%if not %sysfunc(exist(&casdata)) %then %do;
proc casutil;
load data=&sasdata casout="hmeq" outcaslib=casuser;
run;
%end;
/************************************************************************/
/* Explore the data and plot missing values */
/************************************************************************/
proc cardinality data=&casdata outcard=mycaslib.data_card;
run;
proc print data=mycaslib.data_card(where=(_nmiss_>0));
title "Data Summary";
run;
data data_missing;
set mycaslib.data_card (where=(_nmiss_>0) keep=_varname_ _nmiss_ _nobs_);
_percentmiss_ = (_nmiss_/_nobs_)*100;
label _percentmiss_ = 'Percent Missing';
run;
proc sgplot data=data_missing;
title "Percentage of Missing Values";
vbar _varname_ / response=_percentmiss_ datalabel categoryorder=respdesc;
run;
title;
/************************************************************************/
/* Impute missing values */
/************************************************************************/
proc varimpute data=&casdata;
input clage /ctech=mean;
input delinq /ctech=median;
input ninq /ctech=random;
input debtinc yoj /ctech=value cvalues=50,100;
output out=mycaslib.hmeq_prepped copyvars=(_ALL_);
code file="&outdir./impute_score.sas";
run;
/************************************************************************/
/* Identify variables that explain variance in the target */
/************************************************************************/
/* Discriminant analysis for class target */
proc varreduce data=mycaslib.hmeq_prepped technique=discriminantanalysis;
class &target &im_class_inputs.;
reduce supervised &target=&im_class_inputs. &im_interval_inputs. / maxeffects=8;
ods output selectionsummary=summary;
run;
data out_iter (keep=Iteration VarExp Base Increment Parameter);
set summary;
Increment=dif(VarExp);
if Increment=. then Increment=0;
Base=VarExp - Increment;
run;
proc transpose data=out_iter out=out_iter_trans;
by Iteration VarExp Parameter;
run;
proc sort data=out_iter_trans;
label _NAME_='Group';
by _NAME_;
run;
/* Variance explained by Iteration plot */
proc sgplot data=out_iter_trans;
title "Variance Explained by Iteration";
yaxis label="Variance Explained";
vbar Iteration / response=COL1 group=_NAME_;
run;
title;
/************************************************************************/
/* Perform a cluster analysis based on demographic inputs */
/************************************************************************/
proc kclus data=mycaslib.hmeq_prepped standardize=std distance=euclidean maxclusters=6;
input &cluster_inputs. / level=interval;
run;
/************************************************************************/
/* Perform a principal components analysis on the interval valued */
/* input variables */
/************************************************************************/
proc pca data=mycaslib.hmeq_prepped plots=(scree);
var &im_interval_inputs;
run;1 %studio_hide_wrapper;
77
78 /************************************************************************/
79 /* This example illustrates various tools for assaying, assessing, */
80 /* modifying and preparing data prior to modeling. It uses HMEQ */
81 /* dataset as input and produces HMEQ_PREPPED dataset. The HMEQ_PREPPED */
82 /* dataset is used in subsequent examples. */
83 /* */
84 /* The steps include: */
85 /* */
86 /* (1) PREPARE AND EXPLORE */
87 /* a) Load data set into CAS */
88 /* b) Explore */
89 /* c) Impute */
90 /* d) Identify variables that explain variance */
91 /* e) Perform a cluster analysis to identify homogeneous */
92 /* groups in the data */
93 /* f) Perform principal components analysis to assess collineary */
94 /* among candidate, interval valued inputs */
95 /************************************************************************/
96
97 /************************************************************************/
98 /* Setup and initialize for later use in the program */
99 /************************************************************************/
100 /* Define a CAS engine libref for CAS in-memory data tables */
101 libname mycaslib cas caslib=casuser;
NOTE: Libref MYCASLIB was successfully assigned as follows:
Engine: CAS
Physical Name: c871747f-5b0b-2e45-af8f-e24f0fd568ce
102
103 /* Specify the data set names */
104 %let sasdata = sampsio.hmeq;
105 %let casdata = mycaslib.hmeq;
106
107 /* Specify the data set inputs and target */
108 %let class_inputs = reason job;
109 %let interval_inputs = clage clno debtinc loan mortdue value yoj derog delinq ninq;
110 %let target = bad;
111
112 %let im_class_inputs = reason job;
113 %let im_interval_inputs = im_clage clno im_debtinc loan mortdue value im_yoj im_ninq derog im_delinq;
114 %let cluster_inputs = im_clage im_debtinc value;
115
116 /* Specify a folder path to write the temporary output files */
117 %let outdir = &_SASWORKINGDIR;
118
119
120 /************************************************************************/
121 /* Load data into CAS if needed. Data should have been loaded in */
122 /* step 1, it will be loaded here after checking if it exists in CAS */
123 /************************************************************************/
124 %if not %sysfunc(exist(&casdata)) %then %do;
125 proc casutil;
126 load data=&sasdata casout="hmeq" outcaslib=casuser;
127 run;
128 %end;
129
130
131 /************************************************************************/
132 /* Explore the data and plot missing values */
133 /************************************************************************/
134 proc cardinality data=&casdata outcard=mycaslib.data_card;
135 run;
NOTE: Added action set 'cardinality'.
NOTE: Writing cardinality.
NOTE: status = 0.
NOTE: The Cloud Analytic Services server processed the request in 0.935232 seconds.
NOTE: PROCEDURE CARDINALITY used (Total process time):
real time 1.17 seconds
cpu time 0.06 seconds
136
137 proc print data=mycaslib.data_card(where=(_nmiss_>0));
138 title "Data Summary";
139 run;
NOTE: The PROCEDURE PRINT printed page 7.
NOTE: PROCEDURE PRINT used (Total process time):
real time 0.07 seconds
cpu time 0.07 seconds
140
141 data data_missing;
142 set mycaslib.data_card (where=(_nmiss_>0) keep=_varname_ _nmiss_ _nobs_);
143 _percentmiss_ = (_nmiss_/_nobs_)*100;
144 label _percentmiss_ = 'Percent Missing';
145 run;
NOTE: There were 11 observations read from the data set MYCASLIB.DATA_CARD.
WHERE _nmiss_>0;
NOTE: The data set WORK.DATA_MISSING has 11 observations and 4 variables.
NOTE: DATA statement used (Total process time):
real time 0.01 seconds
cpu time 0.01 seconds
146
147 proc sgplot data=data_missing;
148 title "Percentage of Missing Values";
149 vbar _varname_ / response=_percentmiss_ datalabel categoryorder=respdesc;
150 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.27 seconds
cpu time 0.16 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot5.png.
NOTE: There were 11 observations read from the data set WORK.DATA_MISSING.
151 title;
152
153
154 /************************************************************************/
155 /* Impute missing values */
156 /************************************************************************/
157 proc varimpute data=&casdata;
158 input clage /ctech=mean;
159 input delinq /ctech=median;
160 input ninq /ctech=random;
161 input debtinc yoj /ctech=value cvalues=50,100;
162 output out=mycaslib.hmeq_prepped copyvars=(_ALL_);
163 code file="&outdir./impute_score.sas";
164 run;
NOTE: Using SEED=1963189931 for random imputation.
NOTE: External file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/im
pute_score.sas opened.
NOTE: The procedure wrote the DATA step code to external file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/im
pute_score.sas.
NOTE: The Cloud Analytic Services server processed the request in 0.043694 seconds.
NOTE: The data set MYCASLIB.HMEQ_PREPPED has 5960 observations and 18 variables.
NOTE: The PROCEDURE VARIMPUTE printed page 8.
NOTE: PROCEDURE VARIMPUTE used (Total process time):
real time 0.12 seconds
cpu time 0.08 seconds
165
166
167 /************************************************************************/
168 /* Identify variables that explain variance in the target */
169 /************************************************************************/
170 /* Discriminant analysis for class target */
171 proc varreduce data=mycaslib.hmeq_prepped technique=discriminantanalysis;
172 class &target &im_class_inputs.;
173 reduce supervised &target=&im_class_inputs. &im_interval_inputs. / maxeffects=8;
174 ods output selectionsummary=summary;
175 run;
NOTE: The Cloud Analytic Services server processed the request in 0.100386 seconds.
NOTE: The data set WORK.SUMMARY has 8 observations and 9 variables.
NOTE: The PROCEDURE VARREDUCE printed page 9.
NOTE: PROCEDURE VARREDUCE used (Total process time):
real time 0.21 seconds
cpu time 0.06 seconds
176
177 data out_iter (keep=Iteration VarExp Base Increment Parameter);
178 set summary;
179 Increment=dif(VarExp);
180 if Increment=. then Increment=0;
181 Base=VarExp - Increment;
182 run;
NOTE: There were 8 observations read from the data set WORK.SUMMARY.
NOTE: The data set WORK.OUT_ITER has 8 observations and 5 variables.
NOTE: DATA statement used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
183
184 proc transpose data=out_iter out=out_iter_trans;
185 by Iteration VarExp Parameter;
186 run;
NOTE: There were 8 observations read from the data set WORK.OUT_ITER.
NOTE: The data set WORK.OUT_ITER_TRANS has 16 observations and 5 variables.
NOTE: PROCEDURE TRANSPOSE used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
187
188 proc sort data=out_iter_trans;
189 label _NAME_='Group';
190 by _NAME_;
191 run;
NOTE: There were 16 observations read from the data set WORK.OUT_ITER_TRANS.
NOTE: The data set WORK.OUT_ITER_TRANS has 16 observations and 5 variables.
NOTE: PROCEDURE SORT used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
192
193 /* Variance explained by Iteration plot */
194 proc sgplot data=out_iter_trans;
195 title "Variance Explained by Iteration";
196 yaxis label="Variance Explained";
197 vbar Iteration / response=COL1 group=_NAME_;
198 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.22 seconds
cpu time 0.12 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot6.png.
NOTE: There were 16 observations read from the data set WORK.OUT_ITER_TRANS.
199 title;
200
201
202 /************************************************************************/
203 /* Perform a cluster analysis based on demographic inputs */
204 /************************************************************************/
205 proc kclus data=mycaslib.hmeq_prepped standardize=std distance=euclidean maxclusters=6;
206 input &cluster_inputs. / level=interval;
207 run;
NOTE: Using SEED=139918137.
NOTE: Running for input data for K 6.
NOTE: Clustering is finished. MAXITER is satisfied.
NOTE: The Cloud Analytic Services server processed the request in 0.022569 seconds.
NOTE: The PROCEDURE KCLUS printed pages 10-11.
NOTE: PROCEDURE KCLUS used (Total process time):
real time 0.13 seconds
cpu time 0.11 seconds
208
209
210 /************************************************************************/
211 /* Perform a principal components analysis on the interval valued */
212 /* input variables */
213 /************************************************************************/
214 proc pca data=mycaslib.hmeq_prepped plots=(scree);
215 var &im_interval_inputs;
216 run;
WARNING: 1138 observations in table HMEQ_PREPPED omitted due to missing values.
NOTE: The Cloud Analytic Services server processed the request in 0.022168 seconds.
NOTE: The PROCEDURE PCA printed pages 12-14.
NOTE: PROCEDURE PCA used (Total process time):
real time 0.33 seconds
cpu time 0.17 seconds
217
218
219
220
221
222
223 %studio_hide_wrapper;
231
232
/************************************************************************/
/* This example illustrates fitting and comparing several Machine */
/* Learning algorithms for predicting the binary target in the */
/* HMEQ data set. The steps include: */
/* */
/* (1) PREPARE AND EXPLORE */
/* a) Check data is loaded into CAS */
/* */
/* (2) PERFORM SUPERVISED LEARNING */
/* a) Fit a model using a Random Forest */
/* b) Fit a model using Gradient Boosting */
/* c) Fit a model using a Neural Network */
/* d) Fit a model using a Support Vector Machine */
/* */
/* (3) EVALUATE AND IMPLEMENT */
/* a) Score the data */
/* b) Assess model performance */
/* c) Generate ROC and Lift charts */
/************************************************************************/
/************************************************************************/
/* Setup and initialize for later use in the program */
/************************************************************************/
/* Define a CAS engine libref for CAS in-memory data tables */
libname mycaslib cas caslib=casuser;
/* Specify the data set names */
%let casdata = mycaslib.hmeq_prepped;
%let partitioned_data = mycaslib.hmeq_part;
/* Specify the data set inputs and target */
%let class_inputs = reason job;
%let interval_inputs = im_clage clno im_debtinc loan mortdue value im_yoj im_ninq derog im_delinq;
%let target = bad;
/************************************************************************/
/* Check if HMEQ_PREPPED data created in the Prepare and Explore Data */
/* snippet exists. If not, print error message to run the program. */
/************************************************************************/
%if not %sysfunc(exist(&casdata)) %then %do;
%put ERROR: The input dataset HMEQ_PREPPED is not loaded into CAS.;
%put ERROR: Remember to run the Prepare and Explore Data snippet to load necessary data before executing this example.;
%end;
/************************************************************************/
/* Partition the data into training and validation */
/************************************************************************/
proc partition data=&casdata partition samppct=70;
by ⌖
output out=&partitioned_data copyvars=(_ALL_);
run;
/************************************************************************/
/* RANDOM FOREST predictive model */
/************************************************************************/
proc forest data=&partitioned_data ntrees=50 intervalbins=20 minleafsize=5
outmodel=mycaslib.forest_model;
input &interval_inputs. / level = interval;
input &class_inputs. / level = nominal;
target &target / level=nominal;
partition rolevar=_partind_(train='1' validate='0');
run;
/************************************************************************/
/* Score the data using the generated RF model */
/************************************************************************/
proc forest data=&partitioned_data inmodel=mycaslib.forest_model noprint;
output out=mycaslib._scored_RF copyvars=(_ALL_);
run;
/************************************************************************/
/* GRADIENT BOOSTING MACHINES predictive model */
/************************************************************************/
proc gradboost data=&partitioned_data ntrees=10 intervalbins=20 maxdepth=5
outmodel=mycaslib.gb_model;
input &interval_inputs. / level = interval;
input &class_inputs. / level = nominal;
target &target / level=nominal;
partition rolevar=_partind_(train='1' validate='0');
run;
/************************************************************************/
/* Score the data using the generated GBM model */
/************************************************************************/
proc gradboost data=&partitioned_data inmodel=mycaslib.gb_model noprint;
output out=mycaslib._scored_GB copyvars=(_ALL_);
run;
/************************************************************************/
/* NEURAL NETWORK predictive model */
/************************************************************************/
proc nnet data=&partitioned_data;
target &target / level=nom;
input &interval_inputs. / level=int;
input &class_inputs. / level=nom;
hidden 2;
train outmodel=mycaslib.nnet_model;
partition rolevar=_partind_(train='1' validate='0');
ods exclude OptIterHistory;
run;
/************************************************************************/
/* Score the data using the generated NN model */
/************************************************************************/
proc nnet data=&partitioned_data inmodel=mycaslib.nnet_model noprint;
output out=mycaslib._scored_NN copyvars=(_ALL_);
run;
/************************************************************************/
/* SUPPORT VECTOR MACHINE predictive model */
/************************************************************************/
proc svmachine data=&partitioned_data(where=(_partind_=1));
kernel polynom / deg=2;
target ⌖
input &interval_inputs. / level=interval;
input &class_inputs. / level=nominal;
id bad _partind_;
savestate rstore=mycaslib.svm_astore_model;
ods exclude IterHistory;
run;
/************************************************************************/
/* Score data using ASTORE code generated for the SVM model */
/************************************************************************/
proc astore;
score data=&partitioned_data out=mycaslib._scored_SVM
rstore=mycaslib.svm_astore_model;
run;
/************************************************************************/
/* Assess */
/************************************************************************/
%macro assess_model(prefix=, var_evt=, var_nevt=);
proc assess data=mycaslib._scored_&prefix.(where=(_partind_=0));
input &var_evt.;
target &target / level=nominal event='1';
fitstat pvar=&var_nevt. / pevent='0';
ods output
fitstat=&prefix._fitstat
rocinfo=&prefix._rocinfo
liftinfo=&prefix._liftinfo;
run;
%mend assess_model;
ods exclude all;
%assess_model(prefix=RF, var_evt=p_&target.1, var_nevt=p_&target.0);
%assess_model(prefix=SVM, var_evt=p_&target.1, var_nevt=p_&target.0);
%assess_model(prefix=GB, var_evt=p_&target.1, var_nevt=p_&target.0);
%assess_model(prefix=NN, var_evt=p_&target.1, var_nevt=p_&target.0);
ods exclude none;
/************************************************************************/
/* ROC and Lift Charts using validation data */
/************************************************************************/
ods graphics on;
data all_rocinfo;
set SVM_rocinfo(in=s)
GB_rocinfo(in=g)
NN_rocinfo(in=n)
RF_rocinfo(in=f);
length model $ 16;
select;
when (s) model='SVM';
when (f) model='Forest';
when (g) model='GradientBoosting';
when (n) model='NeuralNetwork';
end;
run;
data all_liftinfo;
set SVM_liftinfo(in=s)
GB_liftinfo(in=g)
NN_liftinfo(in=n)
RF_liftinfo(in=f);
length model $ 16;
select;
when (s) model='SVM';
when (f) model='Forest';
when (g) model='GradientBoosting';
when (n) model='NeuralNetwork';
end;
run;
/* Print AUC (Area Under the ROC Curve) */
title "AUC (using validation data) ";
proc sql;
select distinct model, c from all_rocinfo order by c desc;
quit;
/* Draw ROC charts */
proc sgplot data=all_rocinfo aspect=1;
title "ROC Curve (using validation data)";
xaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
yaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
lineparm x=0 y=0 slope=1 / transparency=.7;
series x=fpr y=sensitivity / group=model;
run;
/* Draw lift charts */
proc sgplot data=all_liftinfo;
title "Lift Chart (using validation data)";
yaxis label=' ' grid;
series x=depth y=lift / group=model markers markerattrs=(symbol=circlefilled);
run;
title;
ods graphics off;AUC (using validation data)
model Area Under ROC
Forest 0.926052
GradientBoosting 0.890352
SVM 0.859655
NeuralNetwork 0.514079
/************************************************************************/
/* This example illustrates fitting and comparing two Machine */
/* Learning algorithms for predicting the binary target in the */
/* HMEQ data set. The steps include: */
/* */
/* (1) PREPARE AND EXPLORE */
/* a) Check data is loaded into CAS */
/* */
/* (2) PERFORM SUPERVISED LEARNING */
/* a) Fit model using Logistic Regression */
/* b) Fit a model using a Decision Tree */
/* */
/* (3) EVALUATE AND IMPLEMENT */
/* a) Score the data */
/* b) Assess model performance */
/* c) Generate ROC and Lift charts */
/************************************************************************/
/************************************************************************/
/* Setup and initialize for later use in the program */
/************************************************************************/
/* Define a CAS engine libref for CAS in-memory data tables */
libname mycaslib cas caslib=casuser;
/* Specify the data set names */
%let casdata = mycaslib.hmeq_prepped;
%let partitioned_data = mycaslib.hmeq_part;
/* Specify the data set inputs and target */
%let class_inputs = reason job;
%let interval_inputs = im_clage clno im_debtinc loan mortdue value im_yoj im_ninq derog im_delinq;
%let target = bad;
/* Specify a folder path to write the temporary output files */
%let outdir = &_SASWORKINGDIR;
/************************************************************************/
/* Check if HMEQ_PREPPED data created in the Prepare and Explore Data */
/* snippet exists. If not, print error message to run the program. */
/************************************************************************/
%if not %sysfunc(exist(&casdata)) %then %do;
%put ERROR: The input dataset HMEQ_PREPPED is not loaded into CAS.;
%put ERROR: Remember to run the Prepare and Explore Data snippet to load necessary data before executing this example.;
%end;
/************************************************************************/
/* Partition the data into training and validation */
/************************************************************************/
proc partition data=&casdata partition samppct=70;
by ⌖
output out=&partitioned_data copyvars=(_ALL_);
run;
/************************************************************************/
/* LOGISTIC REGRESSION predictive model */
/************************************************************************/
/* ALL data used for training model */
proc logselect data=&partitioned_data;
class &target &class_inputs.;
model &target.(event='1')=&class_inputs. &interval_inputs.;
selection method=backward;
code file="&outdir./logselect_score.sas" pcatall;
run;
/************************************************************************/
/* Score the data using the generated logistic model score code */
/************************************************************************/
data mycaslib._scored_logistic;
set &partitioned_data;
%include "&outdir./logselect_score.sas";
run;
/************************************************************************/
/* Assess model performance (LOGISTIC REGRESSION) */
/************************************************************************/
ods exclude all;
proc assess data=mycaslib._scored_logistic(where=(_partind_=0));
input p_&target.1;
target &target / level=nominal event='1';
fitstat pvar=p_&target.0/ pevent='0';
ods output fitstat = logit_fitstat
rocinfo = logit_rocinfo
liftinfo = logit_liftinfo;
run;
ods exclude none;
/************************************************************************/
/* DECISION TREE predictive model */
/************************************************************************/
proc treesplit data=&partitioned_data;
input &interval_inputs. / level=interval;
input &class_inputs. / level=nominal;
target &target / level=nominal;
partition rolevar=_partind_(train='1' validate='0');
grow entropy;
prune c45;
code file="&outdir./treeselect_score.sas";
run;
/************************************************************************/
/* Score the data using the generated tree model score code */
/************************************************************************/
data mycaslib._scored_tree;
set &partitioned_data;
%include "&outdir./treeselect_score.sas";
run;
/************************************************************************/
/* Assess tree model performance (DECISTION TREE) */
/************************************************************************/
ods exclude all;
proc assess data=mycaslib._scored_tree(where=(_partind_=0));
input p_&target.1;
target &target / level=nominal event='1';
fitstat pvar=p_&target.0/ pevent='0';
ods output fitstat = tree_fitstat
rocinfo = tree_rocinfo
liftinfo = tree_liftinfo;
run;
ods exclude none;
/*************************************************************************/
/* Create ROC and Lift plots (both models) using validation data */
/*************************************************************************/
ods graphics on;
data all_rocinfo;
set logit_rocinfo(in=l)
tree_rocinfo(in=t);
length model $ 16;
select;
when (l) model='Logistic';
when (t) model='Tree';
end;
run;
data all_liftinfo;
set logit_liftinfo(in=l)
tree_liftinfo(in=t);
length model $ 16;
select;
when (l) model='Logistic';
when (t) model='Tree';
end;
run;
/* Print AUC (Area Under the ROC Curve) */
title "AUC (using validation data)";
proc sql;
select distinct model, c from all_rocinfo order by c desc;
quit;
/* Draw ROC charts */
proc sgplot data=all_rocinfo aspect=1;
title "ROC Curve (using validation data)";
xaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
yaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
lineparm x=0 y=0 slope=1 / transparency=.7;
series x=fpr y=sensitivity / group=model;
run;
/* Draw lift charts */
proc sgplot data=all_liftinfo;
title "Lift Chart (using validation data)";
yaxis label=' ' grid;
series x=depth y=lift / group=model markers markerattrs=(symbol=circlefilled);
run;
title;
ods graphics off;The LOGSELECT Procedure
Parameter Estimates
Parameter DF Estimate Standard
Error Chi-Square Pr > ChiSq
Intercept 1 -7.082694 0.321028 486.7571 <.0001
IM_CLAGE 1 -0.005552 0.000701 62.6437 <.0001
CLNO 1 -0.022179 0.005139 18.6257 <.0001
IM_DEBTINC 1 0.160982 0.006591 596.5662 <.0001
LOAN 1 -0.000017558 0.000005047 12.1040 0.0005
VALUE 1 0.000001266 0.000000981 1.6647 0.1970
IM_NINQ 1 0.122803 0.023566 27.1554 <.0001
DEROG 1 0.600609 0.069674 74.3098 <.0001
IM_DELINQ 1 0.735456 0.054082 184.9287 <.0001
The TREESPLIT Procedure
Variable Importance
Variable Importance Std Dev
Importance Relative
Importance Count
IM_DEBTINC 193.65 0 1.0000 4
IM_DELINQ 37.0480 0 0.1913 5
IM_CLAGE 19.6441 0 0.1014 12
VALUE 9.9259 0 0.0513 11
LOAN 4.7259 0 0.0244 10
REASON 1.0646 0 0.0055 1
JOB -0.1088 0 -56E-5 9
IM_NINQ -0.8289 0 -0.004 4
CLNO -1.0401 0 -0.005 9
DEROG -2.9749 0 -0.015 5
IM_YOJ -8.8294 0 -0.046 16
MORTDUE -9.9475 0 -0.051 8
AUC (using validation data)
model Area Under ROC
Logistic 0.883336
Tree 0.873168
/************************************************************************/
/* This example showcases fitting and assessing Generalized Linear */
/* Models using the GENSELECT Procedure */
/* The steps include: */
/* */
/* (1) PREPARE */
/* a) Check data is loaded into CAS */
/* */
/* (2) Perform Modeling on the Binary Target */
/* a) Assuming binary distribution, using logit link */
/* b) Plot ROC curve */
/************************************************************************/
/************************************************************************/
/* Setup and initialize for later use in the program */
/************************************************************************/
/* Define a CAS engine libref for CAS in-memory data tables */
libname mycaslib cas caslib=casuser;
/* Specify the data set names */
%let casdata = mycaslib.hmeq_prepped;
%let partitioned_data = mycaslib.hmeq_part;
/* Specify the data set inputs and target */
%let class_inputs = reason job;
%let interval_inputs = im_clage clno im_debtinc loan mortdue value im_yoj im_ninq derog im_delinq;
%let target = bad;
/* Specify a folder path to write the temporary output files */
%let outdir = &_SASWORKINGDIR;
/************************************************************************/
/* Check if HMEQ_PREPPED data created in the Prepare and Explore Data */
/* snippet exists. If not, print error message to run the program. */
/************************************************************************/
%if not %sysfunc(exist(&casdata)) %then %do;
%put ERROR: The input dataset HMEQ_PREPPED is not loaded into CAS.;
%put ERROR: Remember to run the Prepare and Explore Data snippet to load necessary data before executing this example.;
%end;
/************************************************************************/
/* Partition the data into training and validation */
/************************************************************************/
proc partition data=&casdata partition samppct=70;
by ⌖
output out=&partitioned_data copyvars=(_ALL_);
run;
/************************************************************************/
/* The binary target model using LOGIT link */
/************************************************************************/
/* Assuming binary distribution, using logit link, fit a GLM using the Genselect procedure */
proc genselect data=&partitioned_data;
class &class_inputs.;
model &target.(event='1')=&interval_inputs. &class_inputs. / dist=binary link=logit;
selection method=forward(select=sbc stop=sbc choose=validate);
partition rolevar=_partind_(train='1' validate='0');
code file="&outdir./glm_model.sas" pcatall;
run;
/************************************************************************/
/* Score the data using the generated GLM score code */
/************************************************************************/
data mycaslib._scored_glm;
set &partitioned_data;
%include "&outdir./glm_model.sas";
run;
/************************************************************************/
/* Assess model performance (GLM) */
/************************************************************************/
ods exclude all;
proc assess data=mycaslib._scored_glm(where=(_partind_=0));
input p_&target.1;
target &target / level=nominal event='1';
fitstat pvar=p_&target.0/ pevent='0';
ods output fitstat = glm_fitstat
rocinfo = glm_rocinfo
liftinfo = glm_liftinfo;
run;
ods exclude none;
/*************************************************************************/
/* Create ROC and Lift plots using validation data */
/*************************************************************************/
ods graphics on;
/* Print AUC (Area Under the ROC Curve) */
title "AUC (using validation data)";
proc sql;
select distinct 'GLM', c from glm_rocinfo order by c desc;
quit;
/* Draw ROC charts */
proc sgplot data=glm_rocinfo aspect=1 noautolegend;
title "ROC Curve (using validation data)";
xaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
yaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
lineparm x=0 y=0 slope=1 / transparency=.7;
series x=fpr y=sensitivity;
run;
/* Draw lift charts */
proc sgplot data=glm_liftinfo;
title "Lift Chart (using validation data)";
yaxis label=' ' grid;
series x=depth y=lift / markers markerattrs=(symbol=circlefilled);
run;
ods graphics off;1 %studio_hide_wrapper;
77
78 /************************************************************************/
79 /* This example showcases fitting and assessing Generalized Linear */
80 /* Models using the GENSELECT Procedure */
81 /* The steps include: */
82 /* */
83 /* (1) PREPARE */
84 /* a) Check data is loaded into CAS */
85 /* */
86 /* (2) Perform Modeling on the Binary Target */
87 /* a) Assuming binary distribution, using logit link */
88 /* b) Plot ROC curve */
89 /************************************************************************/
90
91 /************************************************************************/
92 /* Setup and initialize for later use in the program */
93 /************************************************************************/
94
95 /* Define a CAS engine libref for CAS in-memory data tables */
96 libname mycaslib cas caslib=casuser;
NOTE: Libref MYCASLIB was successfully assigned as follows:
Engine: CAS
Physical Name: c871747f-5b0b-2e45-af8f-e24f0fd568ce
97
98 /* Specify the data set names */
99 %let casdata = mycaslib.hmeq_prepped;
100 %let partitioned_data = mycaslib.hmeq_part;
101
102 /* Specify the data set inputs and target */
103 %let class_inputs = reason job;
104 %let interval_inputs = im_clage clno im_debtinc loan mortdue value im_yoj im_ninq derog im_delinq;
105 %let target = bad;
106
107 /* Specify a folder path to write the temporary output files */
108 %let outdir = &_SASWORKINGDIR;
109
110 /************************************************************************/
111 /* Check if HMEQ_PREPPED data created in the Prepare and Explore Data */
112 /* snippet exists. If not, print error message to run the program. */
113 /************************************************************************/
114 %if not %sysfunc(exist(&casdata)) %then %do;
115 %put ERROR: The input dataset HMEQ_PREPPED is not loaded into CAS.;
116 %put ERROR: Remember to run the Prepare and Explore Data snippet to load necessary data before executing this example.;
117 %end;
118
119
120 /************************************************************************/
121 /* Partition the data into training and validation */
122 /************************************************************************/
123 proc partition data=&casdata partition samppct=70;
124 by ⌖
125 output out=&partitioned_data copyvars=(_ALL_);
126 run;
NOTE: Stratified sampling is in effect.
NOTE: Using SEED=2022380282 for sampling.
NOTE: The Cloud Analytic Services server processed the request in 0.13547 seconds.
NOTE: The data set MYCASLIB.HMEQ_PART has 5960 observations and 19 variables.
NOTE: The PROCEDURE PARTITION printed page 32.
NOTE: PROCEDURE PARTITION used (Total process time):
real time 0.16 seconds
cpu time 0.04 seconds
127
128
129 /************************************************************************/
130 /* The binary target model using LOGIT link */
131 /************************************************************************/
132 /* Assuming binary distribution, using logit link, fit a GLM using the Genselect procedure */
133 proc genselect data=&partitioned_data;
134 class &class_inputs.;
135 model &target.(event='1')=&interval_inputs. &class_inputs. / dist=binary link=logit;
136 selection method=forward(select=sbc stop=sbc choose=validate);
137 partition rolevar=_partind_(train='1' validate='0');
138 code file="&outdir./glm_model.sas" pcatall;
139 run;
NOTE: Convergence criterion (FCONV=1E-7) satisfied.
NOTE: External file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/gl
m_model.sas opened.
NOTE: The procedure wrote the DATA step code to external file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/gl
m_model.sas.
NOTE: The Cloud Analytic Services server processed the request in 0.096725 seconds.
NOTE: The PROCEDURE GENSELECT printed pages 33-35.
NOTE: PROCEDURE GENSELECT used (Total process time):
real time 0.17 seconds
cpu time 0.14 seconds
140
141
142 /************************************************************************/
143 /* Score the data using the generated GLM score code */
144 /************************************************************************/
145 data mycaslib._scored_glm;
146 set &partitioned_data;
147 %include "&outdir./glm_model.sas";
245 run;
NOTE: Running DATA step in Cloud Analytic Services.
NOTE: The DATA step will run in multiple threads.
NOTE: There were 5960 observations read from the table HMEQ_PART in caslib CASUSER(1440420407@qq.com).
NOTE: The table _scored_glm in caslib CASUSER(1440420407@qq.com) has 5960 observations and 22 variables.
NOTE: DATA statement used (Total process time):
real time 0.04 seconds
cpu time 0.00 seconds
246
247
248 /************************************************************************/
249 /* Assess model performance (GLM) */
250 /************************************************************************/
251 ods exclude all;
252 proc assess data=mycaslib._scored_glm(where=(_partind_=0));
253 input p_&target.1;
254 target &target / level=nominal event='1';
255 fitstat pvar=p_&target.0/ pevent='0';
256 ods output fitstat = glm_fitstat
257 rocinfo = glm_rocinfo
258 liftinfo = glm_liftinfo;
259 run;
NOTE: The Cloud Analytic Services server processed the request in 0.008055 seconds.
NOTE: The data set WORK.GLM_LIFTINFO has 21 observations and 21 variables.
NOTE: The data set WORK.GLM_ROCINFO has 101 observations and 21 variables.
NOTE: The data set WORK.GLM_FITSTAT has 1 observations and 6 variables.
NOTE: PROCEDURE ASSESS used (Total process time):
real time 0.02 seconds
cpu time 0.02 seconds
260 ods exclude none;
261
262
263 /*************************************************************************/
264 /* Create ROC and Lift plots using validation data */
265 /*************************************************************************/
266 ods graphics on;
267
268 /* Print AUC (Area Under the ROC Curve) */
269 title "AUC (using validation data)";
270 proc sql;
271 select distinct 'GLM', c from glm_rocinfo order by c desc;
272 quit;
NOTE: The PROCEDURE SQL printed page 36.
NOTE: PROCEDURE SQL used (Total process time):
real time 0.01 seconds
cpu time 0.01 seconds
273
274 /* Draw ROC charts */
275 proc sgplot data=glm_rocinfo aspect=1 noautolegend;
276 title "ROC Curve (using validation data)";
277 xaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
278 yaxis values=(0 to 1 by 0.25) grid offsetmin=.05 offsetmax=.05;
279 lineparm x=0 y=0 slope=1 / transparency=.7;
280 series x=fpr y=sensitivity;
281 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.19 seconds
cpu time 0.12 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot11.png.
NOTE: There were 101 observations read from the data set WORK.GLM_ROCINFO.
282
283 /* Draw lift charts */
284 proc sgplot data=glm_liftinfo;
285 title "Lift Chart (using validation data)";
286 yaxis label=' ' grid;
287 series x=depth y=lift / markers markerattrs=(symbol=circlefilled);
288 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.14 seconds
cpu time 0.09 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot12.png.
NOTE: There were 21 observations read from the data set WORK.GLM_LIFTINFO.
289
290 ods graphics off;
291
292 %studio_hide_wrapper;
300
301
Parameter Estimates
Parameter DF Estimate Standard
Error Chi-Square Pr > ChiSq
Intercept 1 -7.499193 0.392688 364.6975 <.0001
IM_CLAGE 1 -0.004314 0.000837 26.5508 <.0001
CLNO 1 -0.022570 0.006300 12.8341 0.0003
IM_DEBTINC 1 0.169074 0.008056 440.5204 <.0001
LOAN 1 -0.000017455 0.000006161 8.0263 0.0046
MORTDUE 1 -0.000001187 0.000002772 0.1833 0.6686
VALUE 1 0.000001785 0.000002303 0.6007 0.4383
IM_YOJ 1 -0.005669 0.002911 3.7929 0.0515
IM_NINQ 1 0.133431 0.028314 22.2079 <.0001
DEROG 1 0.552326 0.079905 47.7797 <.0001
IM_DELINQ 1 0.722498 0.066444 118.2374 <.0001
AUC (using validation data)
Area Under ROC
GLM 0.876721
/****************************************************************************/
/* This snippet showcases a sample Machine Learning workflow for */
/* unsupervised learning using SASHELP.IRIS data set. The steps include: */
/* */
/* (1) PREPARE DATA */
/* a) Load data set into CAS */
/* */
/* (2) PERFORM UNSUPERVISED LEARNING */
/* a) Generate Principal Components */
/* b) Analyze Clusters */
/* */
/* (3) VISUALIZE THE RESULTS */
/* a) Examine the clustering plot */
/* b) Identify clusters in a PCA plot */
/****************************************************************************/
/****************************************************************************/
/* Define the macro variables for later use in the program */
/****************************************************************************/
/* Specify a folder path to write the temporary output files */
%let outdir = &_SASWORKINGDIR;
/* Create a CAS engine libref to save the output data sets */
%let caslibname = mycas;
libname &caslibname cas caslib=casuser;
/* Specify the data set names */
%let sasdata = sashelp.iris;
%let casdata = &caslibname..iris;
/* Specify the data set inputs */
%let interval_vars=sepallength sepalwidth petallength petalwidth;
/****************************************************************************/
/* Load data into CAS */
/****************************************************************************/
data &casdata;
set &sasdata;
rowid = _n_;
run;
/****************************************************************************/
/* Unsupervised Learning: Principal Component Analysis */
/****************************************************************************/
proc pca data=&casdata prefix=PC method=EIG plots=all;
var &interval_vars;
output out=&casdata._scored_pca copyvars=(_all_) score=PC_;
code file="&outdir/pca.sas";
run;
/****************************************************************************/
/* Unsupervised Learning: Cluster Analysis */
/****************************************************************************/
proc kclus data=&casdata._scored_pca standardize=STD impute=MEAN
distance=EUCLIDEAN maxiters=50 maxclusters=3;
input &interval_vars;
score out=&casdata._scored_kclus copyvars=(_all_);
ods output clustersum=clus_clustersum;
code file="&outdir/cluster.sas";
run;
/****************************************************************************/
/* Visualize the results using a clustering plot for segment frequency */
/****************************************************************************/
data clus_clustersum;
set clus_clustersum;
clusterLabel = catx(' ', 'Cluster', cluster);
run;
proc template;
define statgraph simplepie;
begingraph;
entrytitle "Segment Frequency";
layout region;
piechart category=clusterLabel response=frequency;
endlayout;
endgraph;
end;
run;
proc sgrender data=clus_clustersum template=simplepie;
run;
/****************************************************************************/
/* Visualize the results by identifying clusters in a PCA plot */
/****************************************************************************/
proc sgplot data=&casdata._scored_kclus(keep=PC_1 PC_2 _cluster_id_);
title "Identify Clusters in a PCA Plot";
scatter x=PC_1 y=PC_2 / group=_cluster_id_;
run;
title;1 %studio_hide_wrapper;
77
78 /****************************************************************************/
79 /* This snippet showcases a sample Machine Learning workflow for */
80 /* unsupervised learning using SASHELP.IRIS data set. The steps include: */
81 /* */
82 /* (1) PREPARE DATA */
83 /* a) Load data set into CAS */
84 /* */
85 /* (2) PERFORM UNSUPERVISED LEARNING */
86 /* a) Generate Principal Components */
87 /* b) Analyze Clusters */
88 /* */
89 /* (3) VISUALIZE THE RESULTS */
90 /* a) Examine the clustering plot */
91 /* b) Identify clusters in a PCA plot */
92 /****************************************************************************/
93
94
95 /****************************************************************************/
96 /* Define the macro variables for later use in the program */
97 /****************************************************************************/
98 /* Specify a folder path to write the temporary output files */
99 %let outdir = &_SASWORKINGDIR;
100
101 /* Create a CAS engine libref to save the output data sets */
102 %let caslibname = mycas;
103 libname &caslibname cas caslib=casuser;
NOTE: Libref MYCAS was successfully assigned as follows:
Engine: CAS
Physical Name: c871747f-5b0b-2e45-af8f-e24f0fd568ce
104
105 /* Specify the data set names */
106 %let sasdata = sashelp.iris;
107 %let casdata = &caslibname..iris;
108
109 /* Specify the data set inputs */
110 %let interval_vars=sepallength sepalwidth petallength petalwidth;
111
112 /****************************************************************************/
113 /* Load data into CAS */
114 /****************************************************************************/
115 data &casdata;
116 set &sasdata;
117 rowid = _n_;
118 run;
NOTE: There were 150 observations read from the data set SASHELP.IRIS.
NOTE: The data set MYCAS.IRIS has 150 observations and 6 variables.
NOTE: DATA statement used (Total process time):
real time 0.01 seconds
cpu time 0.00 seconds
119
120 /****************************************************************************/
121 /* Unsupervised Learning: Principal Component Analysis */
122 /****************************************************************************/
123 proc pca data=&casdata prefix=PC method=EIG plots=all;
124 var &interval_vars;
125 output out=&casdata._scored_pca copyvars=(_all_) score=PC_;
126 code file="&outdir/pca.sas";
127 run;
NOTE: External file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/pc
a.sas opened.
NOTE: The procedure wrote the DATA step code to external file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/pc
a.sas.
NOTE: The Cloud Analytic Services server processed the request in 0.013288 seconds.
NOTE: The data set MYCAS.IRIS_SCORED_PCA has 150 observations and 10 variables.
NOTE: The PROCEDURE PCA printed pages 37-38.
NOTE: PROCEDURE PCA used (Total process time):
real time 1.05 seconds
cpu time 0.61 seconds
128
129 /****************************************************************************/
130 /* Unsupervised Learning: Cluster Analysis */
131 /****************************************************************************/
132 proc kclus data=&casdata._scored_pca standardize=STD impute=MEAN
133 distance=EUCLIDEAN maxiters=50 maxclusters=3;
134 input &interval_vars;
135 score out=&casdata._scored_kclus copyvars=(_all_);
136 ods output clustersum=clus_clustersum;
137 code file="&outdir/cluster.sas";
138 run;
NOTE: Using SEED=1870064630.
NOTE: Running for input data for K 3.
NOTE: Clustering is finished. STOPVALUE is satisfied for STOPCRITERION=CLUSTER_CHANGE.
NOTE: External file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/cl
uster.sas opened.
NOTE: The procedure wrote the DATA step code to external file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/cl
uster.sas.
NOTE: The Cloud Analytic Services server processed the request in 0.013027 seconds.
NOTE: The data set WORK.CLUS_CLUSTERSUM has 3 observations and 9 variables.
NOTE: The data set MYCAS.IRIS_SCORED_KCLUS has 150 observations and 13 variables.
NOTE: The PROCEDURE KCLUS printed pages 39-40.
NOTE: PROCEDURE KCLUS used (Total process time):
real time 0.09 seconds
cpu time 0.07 seconds
139
140 /****************************************************************************/
141 /* Visualize the results using a clustering plot for segment frequency */
142 /****************************************************************************/
143 data clus_clustersum;
144 set clus_clustersum;
145 clusterLabel = catx(' ', 'Cluster', cluster);
146 run;
NOTE: There were 3 observations read from the data set WORK.CLUS_CLUSTERSUM.
NOTE: The data set WORK.CLUS_CLUSTERSUM has 3 observations and 10 variables.
NOTE: DATA statement used (Total process time):
real time 0.01 seconds
cpu time 0.01 seconds
147
148 proc template;
149 define statgraph simplepie;
150 begingraph;
151 entrytitle "Segment Frequency";
152 layout region;
153 piechart category=clusterLabel response=frequency;
154 endlayout;
155 endgraph;
156 end;
NOTE: STATGRAPH 'Simplepie' has been saved to: WORK.TEMPLAT
157 run;
NOTE: PROCEDURE TEMPLATE used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
158
159 proc sgrender data=clus_clustersum template=simplepie;
160 run;
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Render.png.
NOTE: There were 3 observations read from the data set WORK.CLUS_CLUSTERSUM.
NOTE: PROCEDURE SGRENDER used (Total process time):
real time 0.18 seconds
cpu time 0.09 seconds
161
162 /****************************************************************************/
163 /* Visualize the results by identifying clusters in a PCA plot */
164 /****************************************************************************/
165 proc sgplot data=&casdata._scored_kclus(keep=PC_1 PC_2 _cluster_id_);
166 title "Identify Clusters in a PCA Plot";
167 scatter x=PC_1 y=PC_2 / group=_cluster_id_;
168 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.18 seconds
cpu time 0.09 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot13.png.
169 title;
170
171 %studio_hide_wrapper;
179
180
/************************************************************************/
/* This snippet showcases a sample Machine Learning workflow for */
/* supervised learning using SAMPLEML.HMEQ data set. The steps include: */
/* */
/* (1) PREPARE AND EXPLORE */
/* a) Load data set into CAS */
/* b) Explore */
/* c) Partition */
/* d) Impute */
/* e) Identify variables that explain variance */
/* */
/* (2) PERFORM SUPERVISED LEARNING */
/* a) Fit model using random forest */
/* */
/* (3) EVALUATE AND IMPLEMENT */
/* a) Score the data */
/* b) Assess model performance */
/* c) Generate ROC and Lift charts */
/************************************************************************/
/************************************************************************/
/* Define the macro variables for later use in the program */
/************************************************************************/
/* Specify a folder path to write the temporary output files */
%let outdir = &_SASWORKINGDIR;
/* Create a CAS engine libref to save the output data sets */
%let caslibname = mycas;
libname &caslibname cas caslib=casuser;
/* Specify the data set names */
%let sasdata = sampsio.hmeq;
%let casdata = &caslibname..hmeq;
%let partitioned_data = &caslibname.._part;
/* Specify the data set inputs and target */
%let class_inputs = reason job derog delinq ninq IM_clno IM_yoj;
%let interval_inputs = loan value IM_clage IM_debtinc IM_mortdue;
%let target = bad;
/************************************************************************/
/* Load data into CAS */
/************************************************************************/
data &casdata;
set &sasdata;
run;
/************************************************************************/
/* Explore the data and look for missing value */
/************************************************************************/
proc cardinality data=&casdata outcard=&caslibname..data_card;
run;
proc print data=&caslibname..data_card(where=(_nmiss_>0));
title "Data Summary";
run;
data data_missing;
set &caslibname..data_card
(where=(_nmiss_>0) keep=_varname_ _nmiss_ _nobs_);
_percentmiss_ = (_nmiss_/_nobs_)*100;
label _percentmiss_ = 'Percent Missing';
run;
proc sgplot data=data_missing;
title "Percentage of Missing Values";
vbar _varname_ / response=_percentmiss_
datalabel categoryorder=respdesc;
run;
title;
/************************************************************************/
/* Partition the data into training and validation */
/************************************************************************/
proc partition data=&casdata partition samppct=70;
by bad;
output out=&partitioned_data copyvars=(_ALL_);
run;
/************************************************************************/
/* Impute missing values */
/************************************************************************/
proc varimpute data=&partitioned_data;
input mortdue yoj clno /ctech=median;
input clage debtinc /ctech=mean;
code file="&outdir./impute1.sas";
output out=&caslibname.._prepped copyvars=(_ALL_);
run;
/************************************************************************/
/* Identify variables that explain variance in the target */
/************************************************************************/
/* Discriminant analysis for class target */
proc varreduce data=&caslibname.._prepped technique=discriminantanalysis;
class &target &class_inputs.;
reduce supervised &target=&class_inputs. &interval_inputs. /
maxeffects=8;
ods output selectionsummary=summary;
run;
data out_iter (keep=Iteration VarExp Base Increment Parameter);
set summary;
Increment=dif(VarExp);
if Increment=. then Increment=0;
Base=VarExp - Increment;
run;
proc transpose data=out_iter out=out_iter_trans;
by Iteration VarExp Parameter;
run;
proc sort data=out_iter_trans;
label _NAME_='Group';
by _NAME_;
run;
/* Variance explained by Iteration plot */
proc sgplot data=out_iter_trans;
title "Variance Explained by Iteration";
yaxis label="Variance Explained";
vbar Iteration / response=COL1 group=_NAME_;
run;
/************************************************************************/
/* Build a predictive model using Random Forest */
/************************************************************************/
proc forest data=&caslibname.._prepped ntrees=50 numbin=20 minleafsize=5;
input &interval_inputs. / level = interval;
input &class_inputs. / level = nominal;
target &target / level = nominal;
partition rolevar=_partind_(train='1' validate='0');
code file="&outdir./forest.sas";
ods output FitStatistics=fitstats;
run;
/************************************************************************/
/* Score the data using the generated model */
/************************************************************************/
data &caslibname.._scored_forest;
set &caslibname.._prepped;
%include "&outdir./forest.sas";
run;
/* create data set from forest stats output */
data fitstats;
set fitstats;
label Trees = 'Number of Trees';
label MiscTrain = 'Training';
label MiscValid = 'Validation';
run;
/* plot misclassification as function of number of trees */
proc sgplot data=fitstats;
title "Training vs Validation";
series x=Trees y=MiscTrain;
series x=Trees y=MiscValid/
lineattrs=(pattern=shortdash thickness=2);
yaxis label='Misclassification Rate';
run;
title;
/************************************************************************/
/* Assess model performance */
/************************************************************************/
proc assess data=&caslibname.._scored_forest;
input p_bad1;
target &target / level=nominal event='1';
fitstat pvar=p_bad0 / pevent='0';
by _partind_;
ods output fitstat = forest_fitstat
rocinfo = forest_rocinfo
liftinfo = forest_liftinfo;
run;
/************************************************************************/
/* Analyze model using ROC and Lift charts */
/************************************************************************/
ods graphics on;
proc format;
value partindlbl 0 = 'Validation' 1 = 'Training';
run;
/* Construct a ROC chart */
proc sgplot data=forest_rocinfo aspect=1;
title "ROC Curve";
xaxis label="False positive rate" values=(0 to 1 by 0.1);
yaxis label="True positive rate" values=(0 to 1 by 0.1);
lineparm x=0 y=0 slope=1 / transparency=.7 LINEATTRS=(Pattern=34);
series x=fpr y=sensitivity /group=_partind_;
format _partind_ partindlbl.;
run;
/* Construct a Lift chart */
proc sgplot data=forest_liftinfo;
title "Lift Chart";
xaxis label="Population Percentage";
yaxis label="Lift";
series x=depth y=lift /
group=_partind_ markers markerattrs=(symbol=circlefilled);
format _partind_ partindlbl.;
run;
title;
ods graphics off;1 %studio_hide_wrapper;
77
78 /************************************************************************/
79 /* This snippet showcases a sample Machine Learning workflow for */
80 /* supervised learning using SAMPLEML.HMEQ data set. The steps include: */
81 /* */
82 /* (1) PREPARE AND EXPLORE */
83 /* a) Load data set into CAS */
84 /* b) Explore */
85 /* c) Partition */
86 /* d) Impute */
87 /* e) Identify variables that explain variance */
88 /* */
89 /* (2) PERFORM SUPERVISED LEARNING */
90 /* a) Fit model using random forest */
91 /* */
92 /* (3) EVALUATE AND IMPLEMENT */
93 /* a) Score the data */
94 /* b) Assess model performance */
95 /* c) Generate ROC and Lift charts */
96 /************************************************************************/
97
98
99 /************************************************************************/
100 /* Define the macro variables for later use in the program */
101 /************************************************************************/
102 /* Specify a folder path to write the temporary output files */
103 %let outdir = &_SASWORKINGDIR;
104
105 /* Create a CAS engine libref to save the output data sets */
106 %let caslibname = mycas;
107 libname &caslibname cas caslib=casuser;
NOTE: Libref MYCAS was successfully assigned as follows:
Engine: CAS
Physical Name: c871747f-5b0b-2e45-af8f-e24f0fd568ce
108
109 /* Specify the data set names */
110 %let sasdata = sampsio.hmeq;
111 %let casdata = &caslibname..hmeq;
112 %let partitioned_data = &caslibname.._part;
113
114 /* Specify the data set inputs and target */
115 %let class_inputs = reason job derog delinq ninq IM_clno IM_yoj;
116 %let interval_inputs = loan value IM_clage IM_debtinc IM_mortdue;
117 %let target = bad;
118
119 /************************************************************************/
120 /* Load data into CAS */
121 /************************************************************************/
122 data &casdata;
123 set &sasdata;
124 run;
NOTE: There were 5960 observations read from the data set SAMPSIO.HMEQ.
NOTE: The data set MYCAS.HMEQ has 5960 observations and 13 variables.
NOTE: DATA statement used (Total process time):
real time 0.01 seconds
cpu time 0.01 seconds
125
126 /************************************************************************/
127 /* Explore the data and look for missing value */
128 /************************************************************************/
129 proc cardinality data=&casdata outcard=&caslibname..data_card;
130 run;
NOTE: Added action set 'cardinality'.
NOTE: Writing cardinality.
NOTE: status = 0.
NOTE: The Cloud Analytic Services server processed the request in 0.705161 seconds.
NOTE: PROCEDURE CARDINALITY used (Total process time):
real time 0.72 seconds
cpu time 0.03 seconds
131
132 proc print data=&caslibname..data_card(where=(_nmiss_>0));
133 title "Data Summary";
134 run;
NOTE: The PROCEDURE PRINT printed page 41.
NOTE: PROCEDURE PRINT used (Total process time):
real time 0.06 seconds
cpu time 0.05 seconds
135
136 data data_missing;
137 set &caslibname..data_card
138 (where=(_nmiss_>0) keep=_varname_ _nmiss_ _nobs_);
139 _percentmiss_ = (_nmiss_/_nobs_)*100;
140 label _percentmiss_ = 'Percent Missing';
141 run;
NOTE: There were 11 observations read from the data set MYCAS.DATA_CARD.
WHERE _nmiss_>0;
NOTE: The data set WORK.DATA_MISSING has 11 observations and 4 variables.
NOTE: DATA statement used (Total process time):
real time 0.01 seconds
cpu time 0.01 seconds
142
143 proc sgplot data=data_missing;
144 title "Percentage of Missing Values";
145 vbar _varname_ / response=_percentmiss_
146 datalabel categoryorder=respdesc;
147 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.19 seconds
cpu time 0.13 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot14.png.
NOTE: There were 11 observations read from the data set WORK.DATA_MISSING.
148 title;
149
150 /************************************************************************/
151 /* Partition the data into training and validation */
152 /************************************************************************/
153 proc partition data=&casdata partition samppct=70;
154 by bad;
155 output out=&partitioned_data copyvars=(_ALL_);
156 run;
NOTE: Stratified sampling is in effect.
NOTE: Using SEED=366064765 for sampling.
NOTE: The Cloud Analytic Services server processed the request in 0.132728 seconds.
NOTE: The data set MYCAS._PART has 5960 observations and 14 variables.
NOTE: The PROCEDURE PARTITION printed page 42.
NOTE: PROCEDURE PARTITION used (Total process time):
real time 0.16 seconds
cpu time 0.02 seconds
157
158 /************************************************************************/
159 /* Impute missing values */
160 /************************************************************************/
161 proc varimpute data=&partitioned_data;
162 input mortdue yoj clno /ctech=median;
163 input clage debtinc /ctech=mean;
164 code file="&outdir./impute1.sas";
165 output out=&caslibname.._prepped copyvars=(_ALL_);
166 run;
NOTE: External file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/im
pute1.sas opened.
NOTE: The procedure wrote the DATA step code to external file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/im
pute1.sas.
NOTE: The Cloud Analytic Services server processed the request in 0.019598 seconds.
NOTE: The data set MYCAS._PREPPED has 5960 observations and 19 variables.
NOTE: The PROCEDURE VARIMPUTE printed page 43.
NOTE: PROCEDURE VARIMPUTE used (Total process time):
real time 0.05 seconds
cpu time 0.02 seconds
167
168 /************************************************************************/
169 /* Identify variables that explain variance in the target */
170 /************************************************************************/
171 /* Discriminant analysis for class target */
172 proc varreduce data=&caslibname.._prepped technique=discriminantanalysis;
173 class &target &class_inputs.;
174 reduce supervised &target=&class_inputs. &interval_inputs. /
175 maxeffects=8;
176 ods output selectionsummary=summary;
177 run;
NOTE: The Cloud Analytic Services server processed the request in 0.059728 seconds.
NOTE: The data set WORK.SUMMARY has 12 observations and 9 variables.
NOTE: The PROCEDURE VARREDUCE printed page 44.
NOTE: PROCEDURE VARREDUCE used (Total process time):
real time 0.10 seconds
cpu time 0.05 seconds
178
179 data out_iter (keep=Iteration VarExp Base Increment Parameter);
180 set summary;
181 Increment=dif(VarExp);
182 if Increment=. then Increment=0;
183 Base=VarExp - Increment;
184 run;
NOTE: There were 12 observations read from the data set WORK.SUMMARY.
NOTE: The data set WORK.OUT_ITER has 12 observations and 5 variables.
NOTE: DATA statement used (Total process time):
real time 0.00 seconds
cpu time 0.01 seconds
185
186 proc transpose data=out_iter out=out_iter_trans;
187 by Iteration VarExp Parameter;
188 run;
NOTE: There were 12 observations read from the data set WORK.OUT_ITER.
NOTE: The data set WORK.OUT_ITER_TRANS has 24 observations and 5 variables.
NOTE: PROCEDURE TRANSPOSE used (Total process time):
real time 0.00 seconds
cpu time 0.01 seconds
189
190 proc sort data=out_iter_trans;
191 label _NAME_='Group';
192 by _NAME_;
193 run;
NOTE: There were 24 observations read from the data set WORK.OUT_ITER_TRANS.
NOTE: The data set WORK.OUT_ITER_TRANS has 24 observations and 5 variables.
NOTE: PROCEDURE SORT used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
194
195 /* Variance explained by Iteration plot */
196 proc sgplot data=out_iter_trans;
197 title "Variance Explained by Iteration";
198 yaxis label="Variance Explained";
199 vbar Iteration / response=COL1 group=_NAME_;
200 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.18 seconds
cpu time 0.09 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot15.png.
NOTE: There were 24 observations read from the data set WORK.OUT_ITER_TRANS.
201
202 /************************************************************************/
203 /* Build a predictive model using Random Forest */
204 /************************************************************************/
205 proc forest data=&caslibname.._prepped ntrees=50 numbin=20 minleafsize=5;
206 input &interval_inputs. / level = interval;
207 input &class_inputs. / level = nominal;
208 target &target / level = nominal;
209 partition rolevar=_partind_(train='1' validate='0');
210 code file="&outdir./forest.sas";
211 ods output FitStatistics=fitstats;
212 run;
NOTE: Using SEED=278020496 for FOREST model building.
NOTE: The Cloud Analytic Services server processed the request in 0.834471 seconds.
NOTE: The Cloud Analytic Services server processed the request in 0.092017 seconds.
NOTE: The Cloud Analytic Services server processed the request in 0.068008 seconds.
NOTE: External file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/fo
rest.sas opened.
NOTE: The procedure wrote the DATA step code to external file
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/fo
rest.sas.
NOTE: The Cloud Analytic Services server processed the request in 0.276402 seconds.
NOTE: The Cloud Analytic Services server processed the request in 0.009015 seconds.
NOTE: The data set WORK.FITSTATS has 50 observations and 10 variables.
NOTE: The PROCEDURE FOREST printed pages 45-46.
NOTE: PROCEDURE FOREST used (Total process time):
real time 5.50 seconds
cpu time 4.28 seconds
213
214 /************************************************************************/
215 /* Score the data using the generated model */
216 /************************************************************************/
217 data &caslibname.._scored_forest;
218 set &caslibname.._prepped;
219 %include "&outdir./forest.sas";
168967 run;
NOTE: Running DATA step in Cloud Analytic Services.
NOTE: The DATA step will run in multiple threads.
NOTE: There were 5960 observations read from the table _PREPPED in caslib CASUSER(1440420407@qq.com).
NOTE: The table _scored_forest in caslib CASUSER(1440420407@qq.com) has 5960 observations and 23 variables.
NOTE: DATA statement used (Total process time):
real time 7.60 seconds
cpu time 1.65 seconds
168968
168969 /* create data set from forest stats output */
168970 data fitstats;
168971 set fitstats;
168972 label Trees = 'Number of Trees';
168973 label MiscTrain = 'Training';
168974 label MiscValid = 'Validation';
168975 run;
NOTE: There were 50 observations read from the data set WORK.FITSTATS.
NOTE: The data set WORK.FITSTATS has 50 observations and 10 variables.
NOTE: DATA statement used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
168976
168977 /* plot misclassification as function of number of trees */
168978 proc sgplot data=fitstats;
168979 title "Training vs Validation";
168980 series x=Trees y=MiscTrain;
168981 series x=Trees y=MiscValid/
168982 lineattrs=(pattern=shortdash thickness=2);
168983 yaxis label='Misclassification Rate';
168984 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.15 seconds
cpu time 0.10 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot16.png.
NOTE: There were 50 observations read from the data set WORK.FITSTATS.
168985 title;
168986
168987 /************************************************************************/
168988 /* Assess model performance */
168989 /************************************************************************/
168990 proc assess data=&caslibname.._scored_forest;
168991 input p_bad1;
168992 target &target / level=nominal event='1';
168993 fitstat pvar=p_bad0 / pevent='0';
168994 by _partind_;
168995 ods output fitstat = forest_fitstat
168996 rocinfo = forest_rocinfo
168997 liftinfo = forest_liftinfo;
168998 run;
NOTE: The Cloud Analytic Services server processed the request in 0.017944 seconds.
NOTE: The data set WORK.FOREST_LIFTINFO has 42 observations and 22 variables.
NOTE: The data set WORK.FOREST_ROCINFO has 202 observations and 22 variables.
NOTE: The data set WORK.FOREST_FITSTAT has 2 observations and 7 variables.
NOTE: The PROCEDURE ASSESS printed pages 47-58.
NOTE: PROCEDURE ASSESS used (Total process time):
real time 0.53 seconds
cpu time 0.54 seconds
168999
169000 /************************************************************************/
169001 /* Analyze model using ROC and Lift charts */
169002 /************************************************************************/
169003 ods graphics on;
169004 proc format;
169005
169005! value partindlbl 0 = 'Validation' 1 = 'Training';
NOTE: Format PARTINDLBL has been output.
169006 run;
NOTE: PROCEDURE FORMAT used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
169007
169008 /* Construct a ROC chart */
169009 proc sgplot data=forest_rocinfo aspect=1;
169010 title "ROC Curve";
169011 xaxis label="False positive rate" values=(0 to 1 by 0.1);
169012 yaxis label="True positive rate" values=(0 to 1 by 0.1);
169013 lineparm x=0 y=0 slope=1 / transparency=.7 LINEATTRS=(Pattern=34);
169014 series x=fpr y=sensitivity /group=_partind_;
169015 format _partind_ partindlbl.;
169016 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.17 seconds
cpu time 0.09 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot17.png.
NOTE: There were 202 observations read from the data set WORK.FOREST_ROCINFO.
169017
169018 /* Construct a Lift chart */
169019 proc sgplot data=forest_liftinfo;
169020 title "Lift Chart";
169021 xaxis label="Population Percentage";
169022 yaxis label="Lift";
169023 series x=depth y=lift /
169024 group=_partind_ markers markerattrs=(symbol=circlefilled);
169025 format _partind_ partindlbl.;
169026 run;
NOTE: PROCEDURE SGPLOT used (Total process time):
real time 0.16 seconds
cpu time 0.10 seconds
NOTE: Listing image output written to
/viya/tmp/ae8ba36b-a659-4da5-bd5e-9dcb33db1425/SAS_workB445000000AE_sas-launcher-10bf8409-d203-431c-a971-7c14500a19cf-6zvff/SG
Plot18.png.
NOTE: There were 42 observations read from the data set WORK.FOREST_LIFTINFO.
169027
169028 title;
169029 ods graphics off;
169030
169031 %studio_hide_wrapper;
169039
169040
Obs _VARNAME_ _FMTWIDTH_ _TYPE_ _RLEVEL_ _ORDER_ _MORE_ _CARDINALITY_ _NOBS_ _SUMFREQS_ _NMISS_ _MISSFMT_ _VISIBLE_ _MIN_ _MAX_ _MEAN_ _STDDEV_ _SKEWNESS_ _KURTOSIS_ _MFREQ_ _MFREQFOUNDLEVEL_ _MFREQNUM_ _MFREQCHR_ _MFREQCFMT_ _LASTNUM_ _LASTCHR_ _LASTCFMT_
1 MORTDUE 12 N INTERVAL ASC Y 20 5960 5442 518 . 9.02 2063 399550 73760.8172 44457.609458 1.8144807024 6.4818663139 5422 N . 5200 5200
2 VALUE 12 N INTERVAL ASC Y 20 5960 5848 112 . 2.21 8000 855909 101776.04874 57385.775334 3.0533442665 24.362804879 5828 N . 16700 16700
3 REASON 7 C CLASS ASC N 2 5960 5708 252 100 . . . . . . 3928 Y . DebtCon DebtCon . HomeImp HomeImp
4 JOB 7 C CLASS ASC N 6 5960 5681 279 100 . . . . . . 2388 Y . Other Other . Self Self
5 YOJ 12 N INTERVAL ASC Y 20 5960 5445 515 . 23.33 0 41 8.9222681359 7.5739822489 0.9884600695 0.3720724789 4569 N . 1.9 1.9
6 DEROG 12 N CLASS ASC N 11 5960 5252 708 . 100 0 10 0.2545696877 0.8460467771 5.3208702503 36.872763388 4527 Y 0 0 10 10
7 DELINQ 12 N CLASS ASC N 14 5960 5380 580 . 100 0 15 0.4494423792 1.1272659176 4.0231495772 23.565448683 4179 Y 0 0 15 15
8 CLAGE 12 N INTERVAL ASC Y 20 5960 5652 308 . 5.55 0 1168.2335609 179.76627519 85.810091764 1.3434120434 7.5995493301 5629 N . 12.176928574 12.176928574
9 NINQ 12 N CLASS ASC N 16 5960 5450 510 . 100 0 17 1.1860550459 1.7286749712 2.6219841723 9.7865072779 2531 Y 0 0 17 17
10 CLNO 12 N INTERVAL ASC Y 20 5960 5738 222 . 49.17 0 71 21.296096201 10.138933192 0.7750517583 1.1576727319 3029 N . 19 19
11 DEBTINC 12 N INTERVAL ASC Y 20 5960 4693 1267 . 21.59 0.5244992154 203.31214869 33.779915349 8.6017461863 2.8523534163 50.504041532 4673 N . 3.7204214203 3.7204214203
