Dataset Thyriod hyperfunction

Basic characteristics Thyriod hyperfunction

191

target objects

One of the thyriod disease databases from UCI containing three classes and about 3800 instances. A patient should be classified as normal, hyper-functioning or subnormal functioning. Class hyperfunctioning is used as target class. Download mat-file with Prtools dataset.

3581

outlier objects

21

features

Unsupervised PCA Thyriod hyperfunction

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 Thyriod hyperfunction

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

Results Thyriod hyperfunction

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

593, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 69.2 ( 0.0) 70.3 ( 0.0) 53.8 ( 0.0)
Min.Cov.Determinant NaN ( 0.0) NaN ( 0.0) 64.1 ( 0.1)
Mixture of Gaussians 64.9 ( 3.1) 68.1 ( 0.9) 53.5 ( 0.7)
Naive Parzen 75.1 ( 0.0) 75.1 ( 0.0) 54.9 ( 0.0)
Parzen 68.3 ( 0.0) 70.6 ( 0.0) 55.9 ( 0.0)
k-means 64.6 ( 1.5) 66.1 ( 2.2) 54.6 ( 1.8)
1-Nearest Neighbors 67.0 ( 0.0) 71.0 ( 0.0) 54.7 ( 0.0)
k-Nearest Neighbors 67.0 ( 0.0) 71.0 ( 0.0) 54.7 ( 0.0)
knn, opt-AUC 67.0 ( 0.0) 66.1 ( 0.6) 54.7 ( 0.0)
Nearest-neighbor dist 60.3 ( 0.1) 67.5 ( 0.0) 52.8 ( 0.0)
Principal comp. 70.0 ( 0.0) 65.9 ( 0.0) 57.2 ( 0.0)
Self-Organ. Map 67.6 ( 0.6) 70.0 ( 0.5) 54.8 ( 1.1)
Auto-enc network 68.3 ( 2.1) 62.8 ( 0.5) 56.5 ( 1.2)
Spanning Tree 67.0 ( 0.0) 71.2 ( 0.0) 52.1 ( 0.0)
L_1-ball 56.2 ( 0.0) 56.2 ( 0.0) 54.3 ( 0.0)
k-center 49.0 ( 2.1) 49.1 ( 1.9) 48.9 ( 2.0)
Support vector DD 51.7 ( 2.6) 45.7 ( 0.0) 51.3 ( 0.0)
Minimax Prob. DD 71.6 ( 0.0) 72.9 ( 0.0) 54.1 ( 0.0)
LinProg DD 48.7 ( 0.0) 52.0 ( 0.0) 51.5 ( 0.0)

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