Classifier: Support Vector Machine (SVM Light)

Corpus One ( date=1950-2006 keywords=american verbose="on"): 394 documents (count greater than 5000 words).
Corpus Two ( date=1950-2006 keywords=NOT+american verbose="on"): 303 documents(count greater than 5000 words).

User EXCLUDE Feature List Submitted ... generating user exclude feature list from input pattern:
color colour honor honour center centre favor favour

Found 9 Features to exclude:

ceńter center centre color colour favor favour honor honour

Run name: bldr

Reading raw data files for Corpus One.
Reading raw data files for Corpus Two.

Pruning Instance Feature Vectors.....
Excluding features found in more than 557 and less than 139....
From 95494 global features, removed 390 found in more than 557 documents, 93034 in less than 139 documents.
2070 features remaining for this classification task.
Done pruning all document vectors. Retained 563197 features.
Deleted 950161 features in too many or too few documents.

Using 2070 features


Training SVMLight
Command:
/projects/datamining/svm_light/svm_learn -x 1 /projects/datamining/philomine/runs/bldr/svmlight/data/bldr.dat /projects/datamining/philomine/runs/bldr/svmlight/models/bldr.model

Result:

Scanning examples...done
Reading examples into memory...100..200..300..400..500..600..OK. (697 examples read)
Setting default regularization parameter C=0.0000
Optimizing........................................................................................................................…
Optimization finished (28 misclassified, maxdiff=0.00099).
Runtime in cpu-seconds: 0.35
Number of SV: 383 (including 280 at upper bound)
L1 loss: loss=142.22003
Norm of weight vector: |w|=0.06905
Norm of longest example vector: |x|=747.27237
Estimated VCdim of classifier: VCdim<=2663.43943
Computing XiAlpha-estimates...done
Runtime for XiAlpha-estimates in cpu-seconds: 0.01
XiAlpha-estimate of the error: error<=53.66% (rho=1.00,depth=0)
XiAlpha-estimate of the recall: recall=>53.05% (rho=1.00,depth=0)
XiAlpha-estimate of the precision: precision=>52.51% (rho=1.00,depth=0)
Number of kernel evaluations: 21979
Computing leave-one-out(?[0]..................................................................................+)(?[1].............…
Retrain on full problem..........................................................................done.
Leave-one-out estimate of the error: error=10.04%
Leave-one-out estimate of the recall: recall=90.10%
Leave-one-out estimate of the precision: precision=91.97%
Actual leave-one-outs computed:  346 (rho=1.00)
Runtime for leave-one-out in cpu-seconds: 34.53
Writing model file...done