OpenML

JavaScript is required to properly view the contents of this page!

TuningSVMs

active ARFF Public Domain (CC0) Visibility: public Uploaded 17-07-2019 by Rafael Gomes Mantovani
0 likes downloaded by 0 people , 0 total downloads 0 issues 0 downvotes

Issue	#Downvotes for this reason	By

Loading wiki

Help us complete this description Edit

Author: Rafael G. Mantovani, Edesio Alcobaça, André L. D. Rossi, Joaquin Vanschoren, André C. P. L. F. de Carvalho Source: "A meta-learning recommender system for hyperparameter tuning: predicting when tuning improves SVM classifiers" - Information Sciences, volume 501, 2019. Please cite: 10.1016/j.ins.2019.06.005 This is a meta-dataset which describes the SVM hyperparameter tuning problem. The target attribute indicates whether tuning is required or default hyperparameter values are enough to each dataset (row). Targets were defined using a statistical labelling rule comparing the predictive performance of models induced with defaults values and tuned ones. In this version of the dataset, the labelling rule was set with 95% confidence.

91 features

Class (target)	nominal	2 unique values 0 missing
datasets (row identifier)	nominal	156 unique values 0 missing
diff.svm.df.lm	numeric	153 unique values 0 missing
diff.svm.nb	numeric	155 unique values 0 missing
diff.svm.stump	numeric	155 unique values 0 missing
diff.svm.nn	numeric	153 unique values 0 missing
diff.nn.lm	numeric	152 unique values 0 missing
diff.nn.stump	numeric	151 unique values 0 missing
diff.nn.nb	numeric	149 unique values 0 missing
diff.nb.stump	numeric	151 unique values 0 missing
diff.nb.lm	numeric	154 unique values 0 missing
diff.stump.lm	numeric	155 unique values 0 missing
simple.classes	numeric	12 unique values 0 missing
simple.attributes	numeric	51 unique values 0 missing
simple.numeric	numeric	46 unique values 0 missing
simple.nominal	numeric	21 unique values 0 missing
simple.samples	numeric	129 unique values 0 missing
simple.dimensionality	numeric	140 unique values 0 missing
simple.numeric_rate	numeric	32 unique values 0 missing
simple.nominal_rate	numeric	32 unique values 0 missing
simple.symbols_min	numeric	11 unique values 0 missing
simple.symbols_max	numeric	22 unique values 0 missing
simple.symbols_mean	numeric	45 unique values 0 missing
simple.symbols_sd	numeric	40 unique values 0 missing
simple.symbols_sum	numeric	44 unique values 0 missing
simple.class_prob_min	numeric	139 unique values 0 missing
simple.class_prob_max	numeric	141 unique values 0 missing
simple.class_prob_mean	numeric	12 unique values 0 missing
simple.class_prob_sd	numeric	131 unique values 0 missing
statistical.skewness	numeric	136 unique values 0 missing
statistical.skewness_prep	numeric	151 unique values 0 missing
statistical.kurtosis	numeric	136 unique values 0 missing
statistical.kurtosis_prep	numeric	151 unique values 0 missing
statistical.abs_cor	numeric	151 unique values 0 missing
statistical.cancor_1	numeric	144 unique values 0 missing
statistical.fract_1	numeric	57 unique values 0 missing
inftheo.class_entropy	numeric	140 unique values 0 missing
inftheo.normalized_class_entropy	numeric	130 unique values 0 missing
inftheo.attribute_entropy	numeric	149 unique values 0 missing
inftheo.normalized_attribute_entropy	numeric	149 unique values 0 missing
inftheo.joint_entropy	numeric	146 unique values 0 missing
inftheo.mutual_information	numeric	151 unique values 0 missing
inftheo.equivalent_attributes	numeric	151 unique values 0 missing
inftheo.noise_signal_ratio	numeric	151 unique values 0 missing
modelbased.nodes	numeric	24 unique values 0 missing
modelbased.leaves.nodes_per_attribute	numeric	24 unique values 0 missing
modelbased.nodes_per_instance	numeric	97 unique values 0 missing
modelbased.leaf_corrobation	numeric	140 unique values 0 missing
modelbased.level_min	numeric	24 unique values 0 missing
modelbased.level_max	numeric	2 unique values 0 missing
modelbased.level_mean	numeric	9 unique values 0 missing
modelbased.level_sd	numeric	38 unique values 0 missing
modelbased.branch_min	numeric	47 unique values 0 missing
modelbased.branch_max	numeric	4 unique values 0 missing
modelbased.branch_mean	numeric	15 unique values 0 missing
modelbased.branch_sd	numeric	72 unique values 0 missing
modelbased.attribute_min	numeric	78 unique values 0 missing
modelbased.attribute_max	numeric	4 unique values 0 missing
modelbased.attribute_mean	numeric	9 unique values 0 missing
modelbased.attribute_sd	numeric	97 unique values 0 missing
modelbased.NA	numeric	122 unique values 0 missing
landmarking.naive_bayes	numeric	149 unique values 0 missing
landmarking.stump_min	numeric	146 unique values 0 missing
landmarking.stump_max	numeric	146 unique values 0 missing
landmarking.stump_mean	numeric	151 unique values 0 missing
landmarking.stump_sd	numeric	122 unique values 0 missing
landmarking.stump_min_gain	numeric	146 unique values 0 missing
landmarking.stump_random	numeric	145 unique values 0 missing
landmarking.nn_1	numeric	146 unique values 0 missing
dcomp.f1	numeric	148 unique values 0 missing
dcomp.f1v	numeric	152 unique values 0 missing
dcomp.f2	numeric	86 unique values 0 missing
dcomp.f3	numeric	126 unique values 0 missing
dcomp.f4	numeric	122 unique values 0 missing
dcomp.l1	numeric	143 unique values 0 missing
dcomp.l2	numeric	112 unique values 0 missing
dcomp.l3	numeric	51 unique values 0 missing
dcomp.n1	numeric	137 unique values 0 missing
dcomp.n2	numeric	142 unique values 0 missing
dcomp.n3	numeric	126 unique values 0 missing
dcomp.n4	numeric	133 unique values 0 missing
dcomp.t1	numeric	48 unique values 0 missing
dcomp.t2	numeric	145 unique values 0 missing
cnet.edges	numeric	151 unique values 0 missing
cnet.degree	numeric	152 unique values 0 missing
cnet.density	numeric	152 unique values 0 missing
cnet.maxComp	numeric	134 unique values 0 missing
cnet.closeness	numeric	63 unique values 0 missing
cnet.betweenness	numeric	152 unique values 0 missing
cnet.clsCoef	numeric	152 unique values 0 missing
cnet.hubs	numeric	151 unique values 0 missing
cnet.avgPath	numeric	152 unique values 0 missing

Show all 91 features

62 properties

NumberOfInstances

156

Number of instances (rows) of the dataset.

NumberOfFeatures

Number of attributes (columns) of the dataset.

NumberOfClasses

Number of distinct values of the target attribute (if it is nominal).

NumberOfMissingValues

Number of missing values in the dataset.

NumberOfInstancesWithMissingValues

Number of instances with at least one value missing.

NumberOfNumericFeatures

Number of numeric attributes.

NumberOfSymbolicFeatures

Number of nominal attributes.

ClassEntropy

0.95

Entropy of the target attribute values.

MeanNoiseToSignalRatio

An estimate of the amount of irrelevant information in the attributes regarding the class. Equals (MeanAttributeEntropy - MeanMutualInformation) divided by MeanMutualInformation.

Quartile2AttributeEntropy

Second quartile (Median) of entropy among attributes.

Dimensionality

0.58

Number of attributes divided by the number of instances.

MeanNominalAttDistinctValues

Average number of distinct values among the attributes of the nominal type.

Quartile2KurtosisOfNumericAtts

2.57

Second quartile (Median) of kurtosis among attributes of the numeric type.

EquivalentNumberOfAtts

Number of attributes needed to optimally describe the class (under the assumption of independence among attributes). Equals ClassEntropy divided by MeanMutualInformation.

MeanSkewnessOfNumericAtts

2.21

Mean skewness among attributes of the numeric type.

Quartile2MeansOfNumericAtts

0.77

Second quartile (Median) of means among attributes of the numeric type.

MajorityClassPercentage

62.82

Percentage of instances belonging to the most frequent class.

MeanStdDevOfNumericAtts

23393.51

Mean standard deviation of attributes of the numeric type.

Quartile2MutualInformation

Second quartile (Median) of mutual information between the nominal attributes and the target attribute.

MajorityClassSize

Number of instances belonging to the most frequent class.

MinAttributeEntropy

Minimal entropy among attributes.

Quartile2SkewnessOfNumericAtts

1.29

Second quartile (Median) of skewness among attributes of the numeric type.

MaxAttributeEntropy

Maximum entropy among attributes.

MinKurtosisOfNumericAtts

-1.22

Minimum kurtosis among attributes of the numeric type.

PercentageOfBinaryFeatures

1.1

Percentage of binary attributes.

Quartile2StdDevOfNumericAtts

0.36

Second quartile (Median) of standard deviation of attributes of the numeric type.

MaxKurtosisOfNumericAtts

154.95

Maximum kurtosis among attributes of the numeric type.

MinMeansOfNumericAtts

-0.13

Minimum of means among attributes of the numeric type.

PercentageOfInstancesWithMissingValues

Percentage of instances having missing values.

Quartile3AttributeEntropy

Third quartile of entropy among attributes.

MaxMeansOfNumericAtts

779933.33

Maximum of means among attributes of the numeric type.

MinMutualInformation

Minimal mutual information between the nominal attributes and the target attribute.

PercentageOfMissingValues

Percentage of missing values.

Quartile3KurtosisOfNumericAtts

16.52

Third quartile of kurtosis among attributes of the numeric type.

MaxMutualInformation

Maximum mutual information between the nominal attributes and the target attribute.

MinNominalAttDistinctValues

The minimal number of distinct values among attributes of the nominal type.

PercentageOfNumericFeatures

98.9

Percentage of numeric attributes.

Quartile3MeansOfNumericAtts

4.18

Third quartile of means among attributes of the numeric type.

MaxNominalAttDistinctValues

The maximum number of distinct values among attributes of the nominal type.

MinSkewnessOfNumericAtts

-6.06

Minimum skewness among attributes of the numeric type.

PercentageOfSymbolicFeatures

1.1

Percentage of nominal attributes.

Quartile3MutualInformation

Third quartile of mutual information between the nominal attributes and the target attribute.

MaxSkewnessOfNumericAtts

12.43

Maximum skewness among attributes of the numeric type.

MinStdDevOfNumericAtts

Minimum standard deviation of attributes of the numeric type.

Quartile1AttributeEntropy

First quartile of entropy among attributes.

Quartile3SkewnessOfNumericAtts

3.53

Third quartile of skewness among attributes of the numeric type.

MaxStdDevOfNumericAtts

2096900.37

Maximum standard deviation of attributes of the numeric type.

MinorityClassPercentage

37.18

Percentage of instances belonging to the least frequent class.

Quartile1KurtosisOfNumericAtts

0.02

First quartile of kurtosis among attributes of the numeric type.

Quartile3StdDevOfNumericAtts

5.41

Third quartile of standard deviation of attributes of the numeric type.

MeanAttributeEntropy

Average entropy of the attributes.

MinorityClassSize

Number of instances belonging to the least frequent class.

Quartile1MeansOfNumericAtts

0.23

First quartile of means among attributes of the numeric type.

StdvNominalAttDistinctValues

Standard deviation of the number of distinct values among attributes of the nominal type.

MeanKurtosisOfNumericAtts

18.12

Mean kurtosis among attributes of the numeric type.

NumberOfBinaryFeatures

Number of binary attributes.

Quartile1MutualInformation

First quartile of mutual information between the nominal attributes and the target attribute.

MeanMeansOfNumericAtts

8717.73

Mean of means among attributes of the numeric type.

Quartile1SkewnessOfNumericAtts

0.1

First quartile of skewness among attributes of the numeric type.

AutoCorrelation

0.55

Average class difference between consecutive instances.

MeanMutualInformation

Average mutual information between the nominal attributes and the target attribute.

Quartile1StdDevOfNumericAtts

0.22

First quartile of standard deviation of attributes of the numeric type.

Show all 62 properties