Error in PL/SQL generated package

Question

Hello,
With the help of ODM (version 10,2,0,3,1 -Build 479) I created a SVM classification model, which works very fine.
After that, I generated the PL/SQL package, which returns an ORA-06512 in the call into DBMS_DATA_MINING.CREATE_MODEL.
I tried to rebuild the model in the ODM and all worked well.
I am kind of stuck here and don't know what to do next.
Thanks,
Igor

Answer 1

Hi,
hope you had a nice vacation.
As for the code I have a feeling I try to fill a varchar2 with text, larger than declared length. I created a standard PL/SQL package as defined in the tutorial and the error remained. I wonder how come that with the use of odminer the building process is successful while the use of PL/SQL package returns errors?
Thanks,
Igor
CREATE VIEW mining_data_build_v AS
SELECT
a.CUST_ID,
a.CUST_GENDER,
2003-a.CUST_YEAR_OF_BIRTH AGE,
a.CUST_MARITAL_STATUS,
c.COUNTRY_NAME,
a.CUST_INCOME_LEVEL,
b.EDUCATION,
b.OCCUPATION,
b.HOUSEHOLD_SIZE,
b.YRS_RESIDENCE,
b.AFFINITY_CARD,
b.BULK_PACK_DISKETTES,
b.FLAT_PANEL_MONITOR,
b.HOME_THEATER_PACKAGE,
b.BOOKKEEPING_APPLICATION,
b.PRINTER_SUPPLIES,
b.Y_BOX_GAMES,
b.OS_DOC_SET_KANJI
FROM
sh.customers a,
sh.supplementary_demographics b,
sh.countries c
WHERE
a.CUST_ID = b.CUST_ID
AND a.country_id = c.country_id
AND a.cust_id between 101501 and 103000;
using the package
CREATE PACKAGE "DATAMININGACTIVITY9" AUTHID DEFINER AS
PROCEDURE "MINING_DATA_BUI666498540_BA"(case_table IN VARCHAR2 DEFAULT '"DMUSER"."MINING_DATA_BUILD_V"',
additional_table_1 IN VARCHAR2 DEFAULT NULL,
model_name IN VARCHAR2 DEFAULT 'MINING_DATA_B80870_SV',
confusion_matrix_name IN VARCHAR2 DEFAULT '"DM4J$T504449054041_M"',
lift_result_name IN VARCHAR2 DEFAULT '"DM4J$T504449083327_L"',
roc_result_name IN VARCHAR2 DEFAULT '"DM4J$T504449092305_R"',
test_metric_name IN VARCHAR2 DEFAULT '"DM4J$MINING_D51278_TM"',
drop_output IN BOOLEAN DEFAULT FALSE);
END;
CREATE PACKAGE BODY "DATAMININGACTIVITY9" AS
c_long_sql_statement_length CONSTANT INTEGER := 32767;
SUBTYPE SQL_STATEMENT_TYPE IS VARCHAR2(32767);
SUBTYPE LONG_SQL_STATEMENT_TYPE IS DBMS_SQL.VARCHAR2A;
TYPE TABLE_ARRAY is TABLE OF VARCHAR2(62);
TYPE LSTMT_REC_TYPE IS RECORD (
lstmt dbms_sql.VARCHAR2A,
lb BINARY_INTEGER DEFAULT 1,
ub BINARY_INTEGER DEFAULT 0);
TYPE LSTMT_REC_TYPE_ARRAY is TABLE OF LSTMT_REC_TYPE;
TYPE QUERY_ARRAY is TABLE OF SQL_STATEMENT_TYPE;
TYPE TARGET_VALUES_LIST IS TABLE OF VARCHAR2(32);
TYPE VALUE_COUNT_LIST IS TABLE OF NUMBER;
PROCEDURE dump_varchar2a(vc2a dbms_sql.VARCHAR2A) IS
v_str varchar2(32767);
BEGIN
DBMS_OUTPUT.PUT_LINE('dump_varchar2a:');
FOR i IN 1..vc2a.COUNT LOOP
v_str := vc2a(i);
DBMS_OUTPUT.PUT_LINE(v_str);
END LOOP;
END;
PROCEDURE ls_append(
r_lstmt IN OUT NOCOPY LSTMT_REC_TYPE,
p_txt VARCHAR2)
IS
BEGIN
r_lstmt.ub := r_lstmt.ub + 1;
r_lstmt.lstmt(r_lstmt.ub) := p_txt;
END ls_append;
PROCEDURE ls_append(
r_lstmt IN OUT NOCOPY LSTMT_REC_TYPE,
p_txt LSTMT_REC_TYPE) IS
BEGIN
FOR i IN p_txt.lb..p_txt.ub LOOP
r_lstmt.ub := r_lstmt.ub + 1;
r_lstmt.lstmt(r_lstmt.ub) := p_txt.lstmt(i);
END LOOP;
END ls_append;
FUNCTION query_valid(
p_query VARCHAR2) RETURN BOOLEAN
IS
v_is_valid BOOLEAN;
BEGIN
BEGIN
EXECUTE IMMEDIATE p_query;
v_is_valid := TRUE;
EXCEPTION WHEN OTHERS THEN
v_is_valid := FALSE;
END;
RETURN v_is_valid;
END query_valid;
FUNCTION table_exist(
p_table_name VARCHAR2) RETURN BOOLEAN IS
BEGIN
RETURN query_valid('SELECT * FROM ' || dbms_assert.simple_sql_name(p_table_name));
END table_exist;
FUNCTION model_exist(
p_model_name VARCHAR2) RETURN BOOLEAN
IS
v_model_cnt NUMBER;
v_model_exists BOOLEAN := FALSE;
BEGIN
SELECT COUNT(*) INTO v_model_cnt FROM DM_USER_MODELS WHERE NAME = UPPER(p_model_name);
IF v_model_cnt > 0 THEN
v_model_exists := TRUE;
END IF;
--DBMS_OUTPUT.PUT_LINE('model exist: '||v_model_exists);
RETURN v_model_exists;
EXCEPTION WHEN OTHERS THEN
RETURN FALSE;
END model_exist;
PROCEDURE drop_table(
p_table_name VARCHAR2)
IS
v_stmt SQL_STATEMENT_TYPE;
BEGIN
v_stmt := 'DROP TABLE '||dbms_assert.simple_sql_name(p_table_name)||' PURGE';
EXECUTE IMMEDIATE v_stmt;
EXCEPTION WHEN OTHERS THEN
NULL;
--DBMS_OUTPUT.PUT_LINE('Failed drop_table: '||p_table_name);
END drop_table;
PROCEDURE drop_view(
p_view_name VARCHAR2)
IS
v_stmt SQL_STATEMENT_TYPE;
BEGIN
v_stmt := 'DROP VIEW '||dbms_assert.simple_sql_name(p_view_name);
EXECUTE IMMEDIATE v_stmt;
EXCEPTION WHEN OTHERS THEN
NULL;
--DBMS_OUTPUT.PUT_LINE('Failed drop_view: '||p_view_name);
END drop_view;
PROCEDURE drop_model(
p_model_name VARCHAR2)
IS
BEGIN
DBMS_DATA_MINING.DROP_MODEL(p_model_name);
EXCEPTION WHEN OTHERS THEN
NULL;
--DBMS_OUTPUT.PUT_LINE('Failed drop_model: '||p_model_name);
END drop_model;
FUNCTION create_new_temp_table_name(prefix IN VARCHAR2, len IN NUMBER)
RETURN VARCHAR2 IS
v_table_name VARCHAR2(30);
v_seed NUMBER;
BEGIN
dbms_random.seed(SYS_GUID());
v_table_name := 'DM$T' || SUBSTR(prefix, 0, 4) || dbms_random.string(NULL, len-8);
--DBMS_OUTPUT.PUT_LINE('create_new_temp_table_name: '||v_table_name);
RETURN v_table_name;
END create_new_temp_table_name;
FUNCTION create_new_temp_table_name(prefix IN VARCHAR2)
RETURN VARCHAR2 IS
BEGIN
RETURN create_new_temp_table_name(prefix, 30);
END create_new_temp_table_name;
FUNCTION ADD_TEMP_TABLE(tempTables IN OUT NOCOPY TABLE_ARRAY, temp_table IN VARCHAR2) RETURN VARCHAR2 IS
BEGIN
tempTables.EXTEND;
tempTables(tempTables.COUNT) := temp_table;
return temp_table;
END;
PROCEDURE DROP_TEMP_TABLES(tempTables IN OUT NOCOPY TABLE_ARRAY) IS
v_temp VARCHAR2(30);
BEGIN
FOR i IN 1..tempTables.COUNT LOOP
v_temp := tempTables(i);
drop_table(v_temp);
drop_view(v_temp);
tempTables.DELETE(i);
END LOOP;
END;
PROCEDURE CHECK_RESULTS(drop_output IN BOOLEAN,
result_name IN VARCHAR2) IS
BEGIN
-- drop all results if drop = true, otherwise make sure all results don't exist already (raise exception)
IF result_name IS NOT NULL THEN
IF drop_output THEN
drop_table(result_name);
drop_view(result_name);
ELSIF (table_exist(result_name)) THEN
RAISE_APPLICATION_ERROR(-20000, 'Result table exists: '||result_name);
END IF;
END IF;
END;
PROCEDURE CHECK_MODEL(drop_output IN BOOLEAN,
model_name IN VARCHAR2) IS
BEGIN
-- drop all results if drop = true, otherwise make sure all results don't exist already (raise exception)
IF model_name IS NOT NULL THEN
IF drop_output THEN
drop_model(model_name);
ELSIF (model_exist(model_name)) THEN
RAISE_APPLICATION_ERROR(-20001, 'Model exists: '||model_name);
END IF;
END IF;
END;
PROCEDURE create_table_from_query(query IN OUT NOCOPY LSTMT_REC_TYPE)
IS
v_cursor NUMBER;
v_feedback INTEGER;
BEGIN
v_cursor := DBMS_SQL.OPEN_CURSOR;
DBMS_SQL.PARSE(
c => v_cursor,
statement => query.lstmt,
lb => query.lb,
ub => query.ub,
lfflg => FALSE,
language_flag => dbms_sql.native);
v_feedback := DBMS_SQL.EXECUTE(v_cursor);
DBMS_SQL.CLOSE_CURSOR(v_cursor);
EXCEPTION WHEN OTHERS THEN
IF DBMS_SQL.IS_OPEN(v_cursor) THEN
DBMS_SQL.CLOSE_CURSOR(v_cursor);
END IF;
RAISE;
END;
FUNCTION get_row_count(tableName IN VARCHAR2)
RETURN INTEGER IS
v_stmt VARCHAR(100);
qcount INTEGER := 0;
BEGIN
v_stmt := 'SELECT COUNT(*) FROM '|| tableName;
EXECUTE IMMEDIATE v_stmt INTO qcount;
RETURN qcount;
END get_row_count;
PROCEDURE SET_EQUAL_DISTRIBUTION (
counts IN OUT VALUE_COUNT_LIST )
IS
v_minvalue NUMBER := 0;
BEGIN
FOR i IN counts.FIRST..counts.LAST
LOOP
IF ( i = counts.FIRST )
THEN
v_minvalue := counts(i);
ELSIF ( counts(i) > 0 AND v_minvalue > counts(i) )
THEN
v_minvalue := counts(i);
END IF;
END LOOP;
FOR i IN counts.FIRST..counts.LAST
LOOP
counts(i) := v_minvalue;
END LOOP;
END SET_EQUAL_DISTRIBUTION;
PROCEDURE GET_STRATIFIED_DISTRIBUTION (
table_name VARCHAR2,
attribute_name VARCHAR2,
percentage NUMBER,
attr_values IN OUT NOCOPY TARGET_VALUES_LIST,
counts IN OUT NOCOPY VALUE_COUNT_LIST,
counts_sampled IN OUT NOCOPY VALUE_COUNT_LIST )
IS
v_tmp_stmt VARCHAR2(4000);
BEGIN
v_tmp_stmt :=
'SELECT /*+ noparallel(t)*/ ' || attribute_name ||
', count(*), ROUND ( ( count(*) * ' || percentage || ') / 100.0 ) FROM '|| table_name ||
' WHERE ' || attribute_name ||' IS NOT NULL GROUP BY ' || attribute_name;
EXECUTE IMMEDIATE v_tmp_stmt
BULK COLLECT INTO attr_values, counts, counts_sampled;
END GET_STRATIFIED_DISTRIBUTION;
FUNCTION GENERATE_STRATIFIED_SQL (
v_2d_temp_view VARCHAR2,
src_table_name VARCHAR2,
attr_names TARGET_VALUES_LIST,
attribute_name VARCHAR2,
percentage NUMBER,
op VARCHAR2,
equal_distribution IN BOOLEAN DEFAULT FALSE) RETURN LSTMT_REC_TYPE
IS
v_tmp_lstmt LSTMT_REC_TYPE;
attr_values_res TARGET_VALUES_LIST;
counts_res VALUE_COUNT_LIST;
counts_sampled_res VALUE_COUNT_LIST;
tmp_str VARCHAR2(4000);
sample_count PLS_INTEGER;
BEGIN
GET_STRATIFIED_DISTRIBUTION(src_table_name, attribute_name, percentage, attr_values_res, counts_res, counts_sampled_res);
IF ( equal_distribution = TRUE )
THEN
SET_EQUAL_DISTRIBUTION(counts_sampled_res);
END IF;
v_tmp_lstmt.ub := 0; -- initialize
ls_append(v_tmp_lstmt, 'CREATE TABLE ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, '( SELECT ');
FOR i IN attr_names.FIRST..attr_names.LAST
LOOP
IF ( i != attr_names.FIRST )
THEN
ls_append(v_tmp_lstmt,',');
END IF;
ls_append(v_tmp_lstmt, attr_names(i));
END LOOP;
ls_append(v_tmp_lstmt, ' FROM (SELECT /*+ no_merge */ t.*, ROWNUM RNUM FROM ' || src_table_name || ' t) WHERE ' );
FOR i IN attr_values_res.FIRST..attr_values_res.LAST
LOOP
IF ( i != attr_values_res.FIRST )
THEN
tmp_str := ' OR ';
END IF;
IF ( counts_res(i) <= 2 ) THEN
sample_count := counts_res(i);
ELSE
sample_count := counts_sampled_res(i);
END IF;
tmp_str := tmp_str ||
'( ' || attribute_name || ' = ''' || attr_values_res(i) || '''' ||
' AND ORA_HASH(RNUM,(' || counts_res(i) || ' -1),12345) ' || op || sample_count || ') ';
ls_append(v_tmp_lstmt, tmp_str );
END LOOP;
ls_append(v_tmp_lstmt, ') ');
return v_tmp_lstmt;
END GENERATE_STRATIFIED_SQL;
PROCEDURE "MINING_DATA_BUI666498540_BA"(case_table IN VARCHAR2 DEFAULT '"DMUSER"."MINING_DATA_BUILD_V"',
additional_table_1 IN VARCHAR2 DEFAULT NULL,
model_name IN VARCHAR2 DEFAULT 'MINING_DATA_B80870_SV',
confusion_matrix_name IN VARCHAR2 DEFAULT '"DM4J$T504449054041_M"',
lift_result_name IN VARCHAR2 DEFAULT '"DM4J$T504449083327_L"',
roc_result_name IN VARCHAR2 DEFAULT '"DM4J$T504449092305_R"',
test_metric_name IN VARCHAR2 DEFAULT '"DM4J$MINING_D51278_TM"',
drop_output IN BOOLEAN DEFAULT FALSE)
IS
additional_data TABLE_ARRAY := TABLE_ARRAY(
additional_table_1
v_tempTables TABLE_ARRAY := TABLE_ARRAY();
v_2d_view VARCHAR2(30);
v_2d_view_build VARCHAR2(30);
v_2d_view_test VARCHAR2(30);
v_2d_temp_view VARCHAR2(30);
v_txn_views TABLE_ARRAY := TABLE_ARRAY();
v_txn_views_build TABLE_ARRAY := TABLE_ARRAY();
v_txn_views_test TABLE_ARRAY := TABLE_ARRAY();
v_txn_temp_views TABLE_ARRAY := TABLE_ARRAY();
v_case_data SQL_STATEMENT_TYPE := case_table;
v_case_id VARCHAR2(30) := 'DMR$CASE_ID';
v_tmp_lstmt LSTMT_REC_TYPE;
v_target_value VARCHAR2(4000) := '1';
v_num_quantiles NUMBER := 10;
v_build_data VARCHAR2(30);
v_test_data VARCHAR2(30);
v_prior VARCHAR2(30);
v_build_setting VARCHAR2(30);
v_apply_result VARCHAR2(30);
v_build_cm VARCHAR2(30);
v_test_cm VARCHAR2(30);
v_accuracy NUMBER;
v_area_under_curve NUMBER;
v_avg_accuracy NUMBER;
v_predictive_confidence NUMBER;
v_confusion_matrix VARCHAR2(30);
v_gen_caseId BOOLEAN := FALSE;
v_2d_txt_view VARCHAR2(30);
v_content_index VARCHAR2(30);
v_content_index_pref VARCHAR2(30);
v_category_temp_table VARCHAR2(30);
v_term_definitions VARCHAR2(30);
v_term_final_table VARCHAR2(30);
v_term_final_table_index VARCHAR2(30);
v_term_final_table_test VARCHAR2(30);
pragma autonomous_transaction;
BEGIN
CHECK_MODEL(drop_output, model_name);
CHECK_RESULTS(drop_output, test_metric_name);
CHECK_RESULTS(drop_output, confusion_matrix_name);
CHECK_RESULTS(drop_output, lift_result_name);
CHECK_RESULTS(drop_output, roc_result_name);
IF (v_gen_caseId) THEN
v_case_data := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
EXECUTE IMMEDIATE 'CREATE TABLE '||v_case_data||' as SELECT rownum as DMR$CASE_ID, t.* FROM ('||case_table||') t ';
EXECUTE IMMEDIATE 'ALTER TABLE '||v_case_data||' add constraint '||create_new_temp_table_name('PK')||' primary key (DMR$CASE_ID)';
END IF;
----- Start: Input Data Preparation -----
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, 'CREATE VIEW ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, ' ( ');
ls_append(v_tmp_lstmt, 'SELECT "CASE_TABLE"."CUST_ID" as "DMR$CASE_ID", TO_CHAR( "CASE_TABLE"."AFFINITY_CARD") AS "AFFINITY_CARD",
"CASE_TABLE"."AGE" AS "AGE",
TO_CHAR( "CASE_TABLE"."BOOKKEEPING_APPLICATION") AS "BOOKKEEPING_APPLICATION",
TO_CHAR( "CASE_TABLE"."BULK_PACK_DISKETTES") AS "BULK_PACK_DISKETTES",
"CASE_TABLE"."COUNTRY_NAME" AS "COUNTRY_NAME",
"CASE_TABLE"."CUST_GENDER" AS "CUST_GENDER",
"CASE_TABLE"."CUST_INCOME_LEVEL" AS "CUST_INCOME_LEVEL",
"CASE_TABLE"."CUST_MARITAL_STATUS" AS "CUST_MARITAL_STATUS",
"CASE_TABLE"."EDUCATION" AS "EDUCATION",
TO_CHAR( "CASE_TABLE"."FLAT_PANEL_MONITOR") AS "FLAT_PANEL_MONITOR",
TO_CHAR( "CASE_TABLE"."HOME_THEATER_PACKAGE") AS "HOME_THEATER_PACKAGE",
"CASE_TABLE"."HOUSEHOLD_SIZE" AS "HOUSEHOLD_SIZE",
"CASE_TABLE"."OCCUPATION" AS "OCCUPATION",
TO_CHAR( "CASE_TABLE"."OS_DOC_SET_KANJI") AS "OS_DOC_SET_KANJI",
TO_CHAR( "CASE_TABLE"."Y_BOX_GAMES") AS "Y_BOX_GAMES",
"CASE_TABLE"."YRS_RESIDENCE" AS "YRS_RESIDENCE" FROM (' || v_case_data || ') CASE_TABLE ');
ls_append(v_tmp_lstmt, ' ) ');
create_table_from_query(v_tmp_lstmt);
v_2d_view := v_2d_temp_view;
----- End: Input Data Preparation -----
----- Start: Outlier Treatment Transformation -----
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, 'CREATE VIEW ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, ' ( ');
ls_append(v_tmp_lstmt, 'SELECT
"AFFINITY_CARD",
( CASE WHEN "AGE" < -1.3 THEN -1.3
WHEN "AGE" >= -1.3 AND "AGE" <= 79.77 THEN "AGE"
WHEN "AGE" > 79.77 THEN 79.77
end) "AGE"
"BOOKKEEPING_APPLICATION",
"BULK_PACK_DISKETTES",
"COUNTRY_NAME",
"CUST_GENDER",
"CUST_INCOME_LEVEL",
"CUST_MARITAL_STATUS",
"DMR$CASE_ID",
"EDUCATION",
"FLAT_PANEL_MONITOR",
"HOME_THEATER_PACKAGE",
"HOUSEHOLD_SIZE",
"OCCUPATION",
"OS_DOC_SET_KANJI",
"Y_BOX_GAMES",
( CASE WHEN "YRS_RESIDENCE" < -1.7 THEN -1.7
WHEN "YRS_RESIDENCE" >= -1.7 AND "YRS_RESIDENCE" <= 10 THEN "YRS_RESIDENCE"
WHEN "YRS_RESIDENCE" > 10 THEN 10
end) "YRS_RESIDENCE"
FROM ');
ls_append(v_tmp_lstmt, v_2d_view);
ls_append(v_tmp_lstmt, ' ) ');
create_table_from_query(v_tmp_lstmt);
v_2d_view := v_2d_temp_view;
----- End: Outlier Treatment Transformation -----
----- Start: Missing Values Transformation -----
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, 'CREATE VIEW ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, ' ( ');
ls_append(v_tmp_lstmt, 'SELECT
DECODE ( "AGE" , NULL,
39.23831 , "AGE" ) "AGE" ,
DECODE ( "YRS_RESIDENCE" , NULL,
4.128 , "YRS_RESIDENCE" ) "YRS_RESIDENCE" ,
DECODE ( "BOOKKEEPING_APPLICATION" , NULL,
''1'' , "BOOKKEEPING_APPLICATION" ) "BOOKKEEPING_APPLICATION" ,
DECODE ( "BULK_PACK_DISKETTES" , NULL,
''1'' , "BULK_PACK_DISKETTES" ) "BULK_PACK_DISKETTES" ,
DECODE ( "COUNTRY_NAME" , NULL,
''United States of America'' , "COUNTRY_NAME" ) "COUNTRY_NAME" ,
DECODE ( "CUST_GENDER" , NULL,
''M'' , "CUST_GENDER" ) "CUST_GENDER" ,
DECODE ( "CUST_INCOME_LEVEL" , NULL,
''J: 190,000 - 249,999'' , "CUST_INCOME_LEVEL" ) "CUST_INCOME_LEVEL" ,
DECODE ( "CUST_MARITAL_STATUS" , NULL,
''Married'' , "CUST_MARITAL_STATUS" ) "CUST_MARITAL_STATUS" ,
DECODE ( "EDUCATION" , NULL,
''HS-grad'' , "EDUCATION" ) "EDUCATION" ,
DECODE ( "FLAT_PANEL_MONITOR" , NULL,
''1'' , "FLAT_PANEL_MONITOR" ) "FLAT_PANEL_MONITOR" ,
DECODE ( "HOME_THEATER_PACKAGE" , NULL,
''1'' , "HOME_THEATER_PACKAGE" ) "HOME_THEATER_PACKAGE" ,
DECODE ( "HOUSEHOLD_SIZE" , NULL,
''3'' , "HOUSEHOLD_SIZE" ) "HOUSEHOLD_SIZE" ,
DECODE ( "OCCUPATION" , NULL,
''Exec.'' , "OCCUPATION" ) "OCCUPATION" ,
DECODE ( "OS_DOC_SET_KANJI" , NULL,
''0'' , "OS_DOC_SET_KANJI" ) "OS_DOC_SET_KANJI" ,
DECODE ( "Y_BOX_GAMES" , NULL,
''0'' , "Y_BOX_GAMES" ) "Y_BOX_GAMES" ,
"AFFINITY_CARD",
"DMR$CASE_ID"
FROM ');
ls_append(v_tmp_lstmt, v_2d_view);
ls_append(v_tmp_lstmt, ' ) ');
create_table_from_query(v_tmp_lstmt);
v_2d_view := v_2d_temp_view;
----- End: Missing Values Transformation -----
----- Start: Normalize Transformation -----
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, 'CREATE VIEW ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, ' ( ');
ls_append(v_tmp_lstmt, 'SELECT
"BOOKKEEPING_APPLICATION",
"BULK_PACK_DISKETTES",
"COUNTRY_NAME",
"CUST_GENDER",
"CUST_INCOME_LEVEL",
"CUST_MARITAL_STATUS",
"EDUCATION",
"FLAT_PANEL_MONITOR",
"HOME_THEATER_PACKAGE",
"HOUSEHOLD_SIZE",
"OCCUPATION",
"OS_DOC_SET_KANJI",
"Y_BOX_GAMES",
"AFFINITY_CARD",
"DMR$CASE_ID",
LEAST(1, GREATEST(0, (ROUND(("AGE" - 17.0) / (79.77 - 17.0),15) * (1.0 - 0.0) + 0.0))) "AGE",
LEAST(1, GREATEST(0, (ROUND(("YRS_RESIDENCE" - 0.0) / (10.0 - 0.0),15) * (1.0 - 0.0) + 0.0))) "YRS_RESIDENCE"
FROM ');
ls_append(v_tmp_lstmt, v_2d_view);
ls_append(v_tmp_lstmt, ' ) ');
create_table_from_query(v_tmp_lstmt);
v_2d_view := v_2d_temp_view;
----- End: Normalize Transformation -----
----- Start: Stratified Split Transformation -----
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, GENERATE_STRATIFIED_SQL(v_2d_temp_view, v_2d_view, TARGET_VALUES_LIST('"BOOKKEEPING_APPLICATION"',
'"BULK_PACK_DISKETTES"',
'"COUNTRY_NAME"',
'"CUST_GENDER"',
'"CUST_INCOME_LEVEL"',
'"CUST_MARITAL_STATUS"',
'"EDUCATION"',
'"FLAT_PANEL_MONITOR"',
'"HOME_THEATER_PACKAGE"',
'"HOUSEHOLD_SIZE"',
'"OCCUPATION"',
'"OS_DOC_SET_KANJI"',
'"Y_BOX_GAMES"',
'"AFFINITY_CARD"',
'"DMR$CASE_ID"',
'"AGE"',
'"YRS_RESIDENCE"'), '"AFFINITY_CARD"', 60, ' < ' ));
create_table_from_query(v_tmp_lstmt);
v_2d_view_build := v_2d_temp_view;
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, GENERATE_STRATIFIED_SQL(v_2d_temp_view, v_2d_view, TARGET_VALUES_LIST('"BOOKKEEPING_APPLICATION"',
'"BULK_PACK_DISKETTES"',
'"COUNTRY_NAME"',
'"CUST_GENDER"',
'"CUST_INCOME_LEVEL"',
'"CUST_MARITAL_STATUS"',
'"EDUCATION"',
'"FLAT_PANEL_MONITOR"',
'"HOME_THEATER_PACKAGE"',
'"HOUSEHOLD_SIZE"',
'"OCCUPATION"',
'"OS_DOC_SET_KANJI"',
'"Y_BOX_GAMES"',
'"AFFINITY_CARD"',
'"DMR$CASE_ID"',
'"AGE"',
'"YRS_RESIDENCE"'), '"AFFINITY_CARD"', 60, ' >= ' ));
create_table_from_query(v_tmp_lstmt);
v_2d_view_test := v_2d_temp_view;
----- End: Stratified Split Transformation -----
----- Start: Mining Data Preparation -----
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, 'CREATE VIEW ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, ' ( ');
ls_append(v_tmp_lstmt,
'SELECT caseTable."AFFINITY_CARD"
, caseTable."AGE"
, caseTable."BOOKKEEPING_APPLICATION"
, caseTable."BULK_PACK_DISKETTES"
, caseTable."COUNTRY_NAME"
, caseTable."CUST_GENDER"
, caseTable."CUST_INCOME_LEVEL"
, caseTable."CUST_MARITAL_STATUS"
, caseTable."DMR$CASE_ID"
, caseTable."EDUCATION"
, caseTable."FLAT_PANEL_MONITOR"
, caseTable."HOME_THEATER_PACKAGE"
, caseTable."HOUSEHOLD_SIZE"
, caseTable."OCCUPATION"
, caseTable."OS_DOC_SET_KANJI"
, caseTable."Y_BOX_GAMES"
, caseTable."YRS_RESIDENCE"
FROM ('); ls_append(v_tmp_lstmt, v_2d_view_build); ls_append(v_tmp_lstmt, ') caseTable
ls_append(v_tmp_lstmt, ' ) ');
create_table_from_query(v_tmp_lstmt);
v_2d_view_build := v_2d_temp_view;
v_tmp_lstmt.ub := 0; -- initialize
v_2d_temp_view := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ls_append(v_tmp_lstmt, 'CREATE VIEW ');
ls_append(v_tmp_lstmt, v_2d_temp_view);
ls_append(v_tmp_lstmt, ' AS ');
ls_append(v_tmp_lstmt, ' ( ');
ls_append(v_tmp_lstmt,
'SELECT caseTable."AFFINITY_CARD"
, caseTable."AGE"
, caseTable."BOOKKEEPING_APPLICATION"
, caseTable."BULK_PACK_DISKETTES"
, caseTable."COUNTRY_NAME"
, caseTable."CUST_GENDER"
, caseTable."CUST_INCOME_LEVEL"
, caseTable."CUST_MARITAL_STATUS"
, caseTable."DMR$CASE_ID"
, caseTable."EDUCATION"
, caseTable."FLAT_PANEL_MONITOR"
, caseTable."HOME_THEATER_PACKAGE"
, caseTable."HOUSEHOLD_SIZE"
, caseTable."OCCUPATION"
, caseTable."OS_DOC_SET_KANJI"
, caseTable."Y_BOX_GAMES"
, caseTable."YRS_RESIDENCE"
FROM ('); ls_append(v_tmp_lstmt, v_2d_view_test); ls_append(v_tmp_lstmt, ') caseTable
ls_append(v_tmp_lstmt, ' ) ');
create_table_from_query(v_tmp_lstmt);
v_2d_view_test := v_2d_temp_view;
v_build_data := v_2d_view_build;
v_test_data := v_2d_view_test;
----- End: Mining Data Preparation -----
v_prior := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
EXECUTE IMMEDIATE 'CREATE TABLE ' || v_prior || ' (TARGET_VALUE VARCHAR2(4000), PRIOR_PROBABILITY NUMBER)';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_prior || ' VALUES (''0'', 0.25333333333333335)';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_prior || ' VALUES (''1'', 0.7466666666666666)';
COMMIT;
v_build_setting := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
EXECUTE IMMEDIATE 'CREATE TABLE ' || v_build_setting || ' (setting_name VARCHAR2(30), setting_value VARCHAR2(128))';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_build_setting || ' VALUES (''JDMS_TARGET_NAME'', ''"AFFINITY_CARD"'')';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_build_setting || ' VALUES (''SVMS_ACTIVE_LEARNING'', ''SVMS_AL_ENABLE'')';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_build_setting || ' VALUES (''JDMS_FUNCTION_TYPE'', ''CLASSIFICATION'')';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_build_setting || ' VALUES (''ALGO_NAME'', ''ALGO_SUPPORT_VECTOR_MACHINES'')';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_build_setting || ' VALUES (''SVMS_CONV_TOLERANCE'', ''0.0010'')';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_build_setting || ' VALUES (''CLAS_PRIORS_TABLE_NAME'', :priorTable)' USING v_prior;
COMMIT;
-- BUILD MODEL
DBMS_DATA_MINING.CREATE_MODEL(
model_name => model_name,
mining_function => dbms_data_mining.classification,
data_table_name => v_build_data,
case_id_column_name => v_case_id,
target_column_name => 'AFFINITY_CARD',
settings_table_name => v_build_setting);
v_test_cm := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
EXECUTE IMMEDIATE 'CREATE TABLE ' || v_test_cm || ' (actual_target_value VARCHAR2(4000), predicted_target_value VARCHAR2(4000), cost NUMBER)';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_test_cm || ' VALUES (''0'', ''0'', 0.0)';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_test_cm || ' VALUES (''0'', ''1'', 1.0)';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_test_cm || ' VALUES (''1'', ''0'', 1.0)';
EXECUTE IMMEDIATE 'INSERT INTO ' || v_test_cm || ' VALUES (''1'', ''1'', 0.0)';
COMMIT;
-- TEST MODEL
IF (test_metric_name IS NOT NULL) THEN
-- CREATE APPLY RESULT FOR TEST
v_apply_result := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
DBMS_DATA_MINING.APPLY(
model_name => model_name,
data_table_name => v_test_data,
case_id_column_name => v_case_id,
result_table_name => v_apply_result);
EXECUTE IMMEDIATE 'CREATE TABLE ' || test_metric_name || ' (METRIC_NAME VARCHAR2(30), METRIC_VARCHAR_VALUE VARCHAR2(31), METRIC_NUM_VALUE NUMBER)';
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''MODEL_NAME'', :model)' USING model_name;
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''TEST_DATA_NAME'', :test_data)' USING v_test_data;
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''MINING_FUNCTION'', ''CLASSIFICATION'')';
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''TARGET_ATTRIBUTE'', :target)' USING 'AFFINITY_CARD';
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''POSITIVE_TARGET_VALUE'', :target_value)' USING v_target_value;
COMMIT;
IF confusion_matrix_name IS NULL THEN
v_confusion_matrix := ADD_TEMP_TABLE(v_tempTables, create_new_temp_table_name('DM$T'));
ELSE
v_confusion_matrix := confusion_matrix_name;
END IF;
DBMS_DATA_MINING.COMPUTE_CONFUSION_MATRIX (
accuracy => v_accuracy,
apply_result_table_name => v_apply_result,
target_table_name => v_test_data,
case_id_column_name => v_case_id,
target_column_name => 'AFFINITY_CARD',
confusion_matrix_table_name => v_confusion_matrix,
score_column_name => 'PREDICTION',
score_criterion_column_name => 'PROBABILITY',
cost_matrix_table_name => v_test_cm);
-- DBMS_OUTPUT.PUT_LINE('**** MODEL ACCURACY ****: ' || ROUND(accuracy, 4));
IF (confusion_matrix_name IS NOT NULL) THEN
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_NUM_VALUE) VALUES (''ACCURACY'', :accuracy)' USING v_accuracy;
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''CONFUSION_MATRIX_TABLE'', :confusion_matrix_name)' USING confusion_matrix_name;
COMMIT;
-- Average Accuracy
EXECUTE IMMEDIATE '
WITH
a as
(SELECT a.actual_target_value, sum(a.value) recall_total
FROM ' || confusion_matrix_name || ' a
group by a.actual_target_value)
b as
(SELECT count(distinct b.actual_target_value) num_recalls
FROM ' || confusion_matrix_name || ' b)
c as
(SELECT c.actual_target_value, value
FROM ' || confusion_matrix_name || ' c
where actual_target_value = predicted_target_value)
d as
(SELECT sum(c.value/a.recall_total) tot_accuracy
FROM a, c
where a.actual_target_value = c.actual_target_value)
SELECT d.tot_accuracy/b.num_recalls * 100 avg_accuracy
FROM b, d' INTO v_avg_accuracy;
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_NUM_VALUE) VALUES (''AVG_ACCURACY'', :avg_accuracy)' USING v_avg_accuracy;
COMMIT;
END IF;
-- Predictive Confidence
EXECUTE IMMEDIATE '
WITH
a as
(SELECT a.actual_target_value, sum(a.value) recall_total
FROM ' || v_confusion_matrix || ' a
group by a.actual_target_value)
b as
(SELECT count(distinct b.actual_target_value) num_classes
FROM ' || v_confusion_matrix || ' b)
c as
(SELECT c.actual_target_value, value
FROM ' || v_confusion_matrix || ' c
where actual_target_value = predicted_target_value)
d as
(SELECT sum(c.value/a.recall_total) tot_accuracy
FROM a, c
where a.actual_target_value = c.actual_target_value)
SELECT (1 - (1 - d.tot_accuracy/b.num_classes) / GREATEST(0.0001, ((b.num_classes-1)/b.num_classes))) * 100
FROM b, d' INTO v_predictive_confidence;
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_NUM_VALUE) VALUES (''PREDICTIVE_CONFIDENCE'', :predictive_confidence)' USING v_predictive_confidence;
COMMIT;
IF lift_result_name IS NOT NULL AND v_target_value IS NOT NULL THEN
DBMS_DATA_MINING.COMPUTE_LIFT (
apply_result_table_name => v_apply_result,
target_table_name => v_test_data,
case_id_column_name => v_case_id,
target_column_name => 'AFFINITY_CARD',
lift_table_name => lift_result_name,
positive_target_value => v_target_value,
num_quantiles => v_num_quantiles,
cost_matrix_table_name => v_test_cm);
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''LIFT_TABLE'', :lift_result_name)' USING lift_result_name;
COMMIT;
END IF;
IF roc_result_name IS NOT NULL AND v_target_value IS NOT NULL THEN
DBMS_DATA_MINING.COMPUTE_ROC (
roc_area_under_curve => v_area_under_curve,
apply_result_table_name => v_apply_result,
target_table_name => v_test_data,
case_id_column_name => v_case_id,
target_column_name => 'AFFINITY_CARD',
roc_table_name => roc_result_name,
positive_target_value => v_target_value,
score_column_name => 'PREDICTION',
score_criterion_column_name => 'PROBABILITY');
-- DBMS_OUTPUT.PUT_LINE('**** AREA UNDER ROC CURVE ****: ' || area_under_curve );
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_VARCHAR_VALUE) VALUES (''ROC_TABLE'', :roc_result_name)' USING roc_result_name;
EXECUTE IMMEDIATE 'INSERT INTO ' || test_metric_name || ' (METRIC_NAME, METRIC_NUM_VALUE) VALUES (''AREA_UNDER_CURVE'', :v_area_under_curve)' USING v_area_under_curve;
COMMIT;
END IF;
END IF;
DROP_TEMP_TABLES(v_tempTables);
EXCEPTION WHEN OTHERS THEN
DROP_TEMP_TABLES(v_tempTables);
RAISE;
END;
END;
/