Dataset Breast Wisconsin (targetcl. benign)

Basic characteristics Breast Wisconsin (targetcl. benign)

458

target objects

The original database of the Wisconsin Breast Cancer Databases from UCI, containing 699 instances, collected between 1989 and 1991. Missing values have been replaced by -1.000000. Download mat-file with Prtools dataset.

241

outlier objects

9

features

Unsupervised PCA Breast Wisconsin (targetcl. benign)

On the left, the PCA scatterplot is shown, on the right the retained variance for varying number of features.
On the left, the PCA scatterplot is shown of data rescaled to unit variance, on the right the retained variance.

Supervised Fisher Breast Wisconsin (targetcl. benign)

On the left, the Fisher scatterplot is shown, on the right the ROC curve along this direction.

Results Breast Wisconsin (targetcl. benign)

The experiments are performed using dd_tools. A rudimentary explanation of the classifiers is given in the classifier section.

505, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 98.5 (0.1) 98.5 (0.1) 98.7 (0.1)
Min.Cov.Determinant NaN (0.0) NaN (0.0) 98.8 (0.1)
Mixture of Gaussians 98.3 (0.2) 98.4 (0.2) 98.4 (0.1)
Naive Parzen 98.7 (0.1) 98.7 (0.1) 98.7 (0.1)
Parzen 99.2 (0.1) 99.1 (0.1) 99.0 (0.1)
k-means 99.1 (0.1) 98.4 (0.1) 99.0 (0.1)
1-Nearest Neighbors 99.2 (0.1) 99.1 (0.1) 99.0 (0.1)
k-Nearest Neighbors 99.2 (0.1) 99.1 (0.1) 99.0 (0.1)
Nearest-neighbor dist 86.7 (0.8) 85.7 (0.3) 84.3 (0.7)
Principal comp. 95.1 (0.2) 92.0 (0.4) 91.8 (0.6)
Self-Organ. Map 99.1 (0.1) 99.0 (0.2) 99.1 (0.1)
Auto-enc network 96.4 (0.2) 96.0 (0.2) 95.1 (0.2)
MST 99.1 (0.1) 99.2 (0.1) 99.0 (0.1)
L_1-ball 98.7 (0.1) 98.0 (0.1) 95.7 (0.7)
k-center 98.7 (0.2) 98.4 (0.2) 98.6 (0.4)
Support vector DD 99.0 (0.1) 98.8 (0.1) 98.9 (0.1)
Minimax Prob. DD 99.2 (0.1) 99.1 (0.1) 99.0 (0.1)
LinProg DD 99.3 (0.1) 98.9 (0.1) 99.2 (0.1)
Lof DD 61.7 (1.0) 72.3 (0.4) 64.1 (1.4)
Lof range DD 34.7 (0.8) 51.5 (1.4) 55.6 (0.9)
Loci DD 99.3 (0.1) 98.7 (0.1) 99.1 (0.1)

Classifier projection spaces The first classifier projection spaces are obtained by computing the classifier label disagreements (setting the threshold on 10% target error) and applying an MDS on the resulting distance matrix between classifiers:



Original



Unit variance



PCA mapped

Classifier projection spaces The second versions of the classifier projection spaces are obtained by computing the classifier ranking disagreements and applying an MDS on the resulting distance matrix between classifiers:



Original



Unit variance



PCA mapped