1. TITLE:
Australian Credit Approval
2. USE IN STATLOG
2.1- Testing Mode
10-Fold Cross Validation
2.2- Special Preprocessing
Yes (See REMARKS)
2.3- Test Results
Error Rate TIME
Algorithm Train Test Train Test
--------- ----------------------------
Cal5 0.132 0.131 24 2
Itrule 0.162 0.137 174 1
LogDisc 0.125 0.141 21 18
Discrim 0.139 0.141 32 7
Dipol92 0.139 0.141 56 2
Radial 0.107 0.145 12 2
Cart 0.145 0.145 68 2
Castle 0.144 0.148 47 5
Bayes 0.136 0.151 4 0
IndCart 0.081 0.152 34 33
BackProp 0.087 0.154 1370 0
C4.5 0.099 0.155 6 1
Smart 0.090 0.158 246 0
BayTree 0 0.171 7 0
KNN 0 0.181 3 7
Ac2 0 0.181 400 14
NewId 0 0.181 15 0
LVQ 0.065 0.197 261 7
Alloc80 0.194 0.201 877 ?
Cn2 0.001 0.204 42 3
QuaDisc 0.185 0.207 31 7
Default 0.440 0.440 ? ?
Cascade 1.000
Kohonen 1.000
3. SOURCES and PAST USAGE
3.1 ORIGINAL SOURCE
(confidential)
Submitted by quinlan@cs.su.oz.au
3.2 PAST USAGE
See Quinlan,
* "Simplifying decision trees", Int J Man-Machine Studies 27,
Dec 1987, pp. 221-234.
* "C4.5: Programs for Machine Learning", Morgan Kaufmann, Oct 1992
3.2. RELEVANT INFORMATION
This file concerns credit card applications. All attribute names
and values have been changed to meaningless symbols to protect
confidentiality of the data.
This dataset is interesting because there is a good mix of
attributes -- continuous, nominal with small numbers of
values, and nominal with larger numbers of values. There
were originally a few missing values, but these have all
been replaced by the overall median.
4. DATASET DESCRIPTION
NUMBER OF EXAMPLES
Total no. = 690
NUMBER OF CLASSES: 2
0,1 (-,+)
Class Distribution:
+: 307 (44.5%) CLASS 1
-: 383 (55.5%) CLASS 0
NUMBER OF ATTRIBUTES
14 (6 Continuous 8 Categorical)
A1: 0,1 CATEGORICAL
a,b
A2: continuous.
A3: continuous.
A4: 1,2,3 CATEGORICAL
p,g,gg
A5: 1, 2,3,4,5, 6,7,8,9,10,11,12,13,14 CATEGORICAL
ff,d,i,k,j,aa,m,c,w, e, q, r,cc, x
A6: 1, 2,3, 4,5,6,7,8,9 CATEGORICAL
ff,dd,j,bb,v,n,o,h,z
A7: continuous.
A8: 1, 0 CATEGORICAL
t, f.
A9: 1, 0 CATEGORICAL
t, f.
A10: continuous.
A11: 1, 0 CATEGORICAL
t, f.
A12: 1, 2, 3 CATEGORICAL
s, g, p
A13: continuous.
A14: continuous.
5-REMARKS:
Missing Attribute Values:
37 cases (5%) HAD one or more missing values. The missing
values from particular attributes WERE:
A1: 12
A2: 12
A4: 6
A5: 6
A6: 9
A7: 9
A14: 13
THESE WERE REPLACED BY THE MODE OF THE ATTRIBUTE (CATEGORICAL)
MEAN OF THE ATTRIBUTE (CONTINUOUS)
There is no cost matrix.
_____________________________________________________________________________
Three remarks relating to the StatLog version:
THE LABELS HAVE BEEN CHANGED FOR THE CONVENIENCE OF THE STATISTICAL
ALGORITHMS. FOR EXAMPLE, ATTRIBUTE 4 ORIGINALLY HAD 3 LABELS p,g,gg
AND THESE HAVE BEEN CHANGED TO LABELS 1,2,3.
1. Attributes 4 and 5 of the original WERE APPARENTLY IDENTICAL,
so ATTRIBUTE 4 OF THE ORIGINAL WAS REMOVED
(for the convenience of the statistical algorithms).
2. Where attributes were categorical, the categories were given numerical
labels in the order of the relative risk of being class "+". Treat
as categorical if thought desirable. All StatLog trials treated
these variables as numerical.
3. A stepwise regression procedure strongly suggests that only
attributes A5, A8, A9, A13 and A14 are relevant. Improved results
are often obtained if only these five attributes are used.
CONTACTS
statlog-adm@ncc.up.pt
bob@stams.strathclyde.ac.uk
See README file for general information
================================================================================
Australian Credit Approval
2. USE IN STATLOG
2.1- Testing Mode
10-Fold Cross Validation
2.2- Special Preprocessing
Yes (See REMARKS)
2.3- Test Results
Error Rate TIME
Algorithm Train Test Train Test
--------- ----------------------------
Cal5 0.132 0.131 24 2
Itrule 0.162 0.137 174 1
LogDisc 0.125 0.141 21 18
Discrim 0.139 0.141 32 7
Dipol92 0.139 0.141 56 2
Radial 0.107 0.145 12 2
Cart 0.145 0.145 68 2
Castle 0.144 0.148 47 5
Bayes 0.136 0.151 4 0
IndCart 0.081 0.152 34 33
BackProp 0.087 0.154 1370 0
C4.5 0.099 0.155 6 1
Smart 0.090 0.158 246 0
BayTree 0 0.171 7 0
KNN 0 0.181 3 7
Ac2 0 0.181 400 14
NewId 0 0.181 15 0
LVQ 0.065 0.197 261 7
Alloc80 0.194 0.201 877 ?
Cn2 0.001 0.204 42 3
QuaDisc 0.185 0.207 31 7
Default 0.440 0.440 ? ?
Cascade 1.000
Kohonen 1.000
3. SOURCES and PAST USAGE
3.1 ORIGINAL SOURCE
(confidential)
Submitted by quinlan@cs.su.oz.au
3.2 PAST USAGE
See Quinlan,
* "Simplifying decision trees", Int J Man-Machine Studies 27,
Dec 1987, pp. 221-234.
* "C4.5: Programs for Machine Learning", Morgan Kaufmann, Oct 1992
3.2. RELEVANT INFORMATION
This file concerns credit card applications. All attribute names
and values have been changed to meaningless symbols to protect
confidentiality of the data.
This dataset is interesting because there is a good mix of
attributes -- continuous, nominal with small numbers of
values, and nominal with larger numbers of values. There
were originally a few missing values, but these have all
been replaced by the overall median.
4. DATASET DESCRIPTION
NUMBER OF EXAMPLES
Total no. = 690
NUMBER OF CLASSES: 2
0,1 (-,+)
Class Distribution:
+: 307 (44.5%) CLASS 1
-: 383 (55.5%) CLASS 0
NUMBER OF ATTRIBUTES
14 (6 Continuous 8 Categorical)
A1: 0,1 CATEGORICAL
a,b
A2: continuous.
A3: continuous.
A4: 1,2,3 CATEGORICAL
p,g,gg
A5: 1, 2,3,4,5, 6,7,8,9,10,11,12,13,14 CATEGORICAL
ff,d,i,k,j,aa,m,c,w, e, q, r,cc, x
A6: 1, 2,3, 4,5,6,7,8,9 CATEGORICAL
ff,dd,j,bb,v,n,o,h,z
A7: continuous.
A8: 1, 0 CATEGORICAL
t, f.
A9: 1, 0 CATEGORICAL
t, f.
A10: continuous.
A11: 1, 0 CATEGORICAL
t, f.
A12: 1, 2, 3 CATEGORICAL
s, g, p
A13: continuous.
A14: continuous.
5-REMARKS:
Missing Attribute Values:
37 cases (5%) HAD one or more missing values. The missing
values from particular attributes WERE:
A1: 12
A2: 12
A4: 6
A5: 6
A6: 9
A7: 9
A14: 13
THESE WERE REPLACED BY THE MODE OF THE ATTRIBUTE (CATEGORICAL)
MEAN OF THE ATTRIBUTE (CONTINUOUS)
There is no cost matrix.
_____________________________________________________________________________
Three remarks relating to the StatLog version:
THE LABELS HAVE BEEN CHANGED FOR THE CONVENIENCE OF THE STATISTICAL
ALGORITHMS. FOR EXAMPLE, ATTRIBUTE 4 ORIGINALLY HAD 3 LABELS p,g,gg
AND THESE HAVE BEEN CHANGED TO LABELS 1,2,3.
1. Attributes 4 and 5 of the original WERE APPARENTLY IDENTICAL,
so ATTRIBUTE 4 OF THE ORIGINAL WAS REMOVED
(for the convenience of the statistical algorithms).
2. Where attributes were categorical, the categories were given numerical
labels in the order of the relative risk of being class "+". Treat
as categorical if thought desirable. All StatLog trials treated
these variables as numerical.
3. A stepwise regression procedure strongly suggests that only
attributes A5, A8, A9, A13 and A14 are relevant. Improved results
are often obtained if only these five attributes are used.
CONTACTS
statlog-adm@ncc.up.pt
bob@stams.strathclyde.ac.uk
See README file for general information
================================================================================
Post a Comment