Dataset Arrhythmia normal (targetcl. 1)

Basic characteristics Arrhythmia normal (targetcl. 1)

183

target objects

The Arrhymthmia database from UCI. The aim is to distinguish between the presence and absence of cardiac arrhythmia and to classify it in one of the 16 groups.Entries with missing values have been removed. Download mat-file with Prtools dataset.

237

outlier objects

278

features

Unsupervised PCA Arrhythmia normal (targetcl. 1)

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 Arrhythmia normal (targetcl. 1)

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

Results Arrhythmia normal (targetcl. 1)

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

515, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 28.7 (0.7) 30.3 (0.5) 25.7 (0.2)
Min.Cov.Determinant NaN (0.0) NaN (0.0) 25.7 (0.4)
Mixture of Gaussians 42.3 (16.6) 30.3 (0.8) 25.8 (0.4)
Naive Parzen 31.4 (0.6) 31.4 (0.6) 25.6 (0.4)
Parzen 42.3 (16.6) 31.0 (0.3) 28.4 (0.7)
k-means 26.2 (0.6) 30.0 (1.0) 26.8 (0.6)
1-Nearest Neighbors 27.9 (0.7) 31.0 (0.3) 29.2 (0.7)
k-Nearest Neighbors 27.9 (0.7) 31.0 (0.3) 29.2 (0.7)
Nearest-neighbor dist 43.3 (1.4) 35.7 (0.9) 46.4 (1.1)
Principal comp. 27.1 (0.6) 29.7 (0.6) 30.2 (0.8)
Self-Organ. Map 25.6 (0.5) 30.7 (0.8) 26.3 (1.0)
Auto-enc network NaN (0.0) NaN (0.0) 29.1 (0.9)
MST 28.3 (0.3) 31.4 (0.5) 29.8 (0.5)
L_1-ball 42.0 (16.5) 42.0 (16.5) 29.9 (1.0)
k-center 29.3 (1.7) 33.8 (0.8) 29.3 (1.5)
Support vector DD 41.9 (16.4) 52.2 (9.1) 48.7 (7.5)
Minimax Prob. DD 0.0 (0.0) 31.0 (0.2) 0.0 (0.0)
LinProg DD 42.3 (16.6) 30.1 (0.6) 42.3 (16.6)
Lof DD 32.2 (1.0) 33.0 (0.6) 32.8 (1.1)
Lof range DD 26.7 (0.3) 31.3 (0.6) 26.7 (0.6)
Loci DD 26.9 (0.7) 30.8 (0.5) 27.5 (0.8)

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