[MADLIB-605] SVM Regression: Accurancy should be improved for some data sets

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Abandoned
• Fix Version/s: v1.9
• Labels: severity_set
• Assignee: Rahul Iyer
• Reporter: Jiali Yao
• Created: 30/Jul/12 15:22
• Updated: 22/Mar/18 18:46
• Resolved: 22/Mar/18 18:46

Description

We run comparable test cases both in MADlib and libsvm and compared mean square error.
We found that below data sets have worse score than svm

1. Kernel function is dot:

Data Sets	MADlib(Parallel = true)	MADlib(Parallel = false)	libsvm	        Madlib/libsvm
bodyfat	        249962.1847	        4397613.616             	4.68E-05	5336898594
mpg	        239380954.3	        1.89706E+11	                22.5239	        10627864.37

Test case:
SELECT madlib.svm_regression
                        ( 'madlibtestdata.svm_bodyfat'::text     --input_table
                        , 'madlibtestresult.reg_model_table'::text    --model_table
                        , 'true'::boolean       --parallel
                        , 'madlib.svm_dot'::text    --kernel_func
                        , 'false'::boolean        --verbose
                        , '0.1'::float8            --eta
                        , '0.005'::float8             --nu
                        , '0.05'::float8        --slambda
                   ) AS q;

SELECT madlib.svm_regression
                        ( 'madlibtestdata.svm_mpg'::text     --input_table
                        , 'madlibtestresult.reg_model_table'::text    --model_table
                        , 'true'::boolean       --parallel
                        , 'madlib.svm_dot'::text    --kernel_func
                        , 'false'::boolean        --verbose
                        , '0.1'::float8            --eta
                        , '0.005'::float8             --nu
                        , '0.05'::float8        --slambda
                   ) AS q;

2. Polynomial

Data Sets	MADlib(Parallel = true)	MADlib(Parallel = false)	libsvm	Madlib/libsvm
bodyfat	        4.07E+26	1.86E+27	0.00143458	2.83446E+29
cpusmall	2.38E+71	4.41E+72	1.42E+42	1.67986E+29
housing  	9.31E+29	6.79E+31	249267	        3.73671E+24
mpg	        2.25E+37	9.89E+39	610.474	        3.68346E+34

Test case example:
SELECT madlib.svm_regression
                        ( 'madlibtestdata.svm_bodyfat'::text     --input_table
                        , 'madlibtestresult.reg_model_table'::text    --model_table
                        , 'true'::boolean       --parallel
                        , 'madlibtestdata.svm_polynomial'::text    --kernel_func
                        , 'false'::boolean        --verbose
                        , '0.1'::float8            --eta
                        , '0.005'::float8             --nu
                        , '0.05'::float8        --slambda
                   ) AS q;

3. Data sets

Data Sets Name	TrainSize	Attr	URL
 abalone 	 4,177 	 8 	 http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/regression.html#abalone 
 bodyfat 	 252 	 14 	 http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/regression.html#bodyfat 
 cpusmall 	 8,192 	 12 	 http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/regression.html#cpusmalll 
 housing 	 506 	 13 	 http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/regression.html#housing