Dataset Vehicle saab

Basic characteristics Vehicle saab

217

target objects

The vehicle dataset from Statlog, to recognize a vehicle from its silhouette. Class saab is used as target class. Download mat-file with Prtools dataset.

629

outlier objects

18

features

Unsupervised PCA Vehicle saab

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 Vehicle saab

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

Results Vehicle saab

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

613, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 79.4 ( 0.1) 87.3 ( 0.2) 56.4 ( 0.1)
Min.Cov.Determinant 86.1 ( 0.3) 86.1 ( 0.3) 62.2 ( 0.5)
Mixture of Gaussians 74.6 ( 0.5) 79.3 ( 2.3) 54.6 ( 0.8)
Naive Parzen 68.4 ( 0.5) 68.4 ( 0.5) 52.6 ( 0.5)
Parzen 60.6 ( 1.9) 72.7 ( 1.8) 52.6 ( 0.5)
k-means 64.8 ( 2.5) 74.9 ( 0.8) 53.6 ( 3.3)
1-Nearest Neighbors 49.2 ( 3.1) 77.5 ( 0.3) 57.6 ( 1.0)
k-Nearest Neighbors 49.2 ( 3.1) 77.5 ( 0.3) 57.6 ( 1.0)
knn, opt-AUC 50.6 ( 4.4) 77.5 ( 0.3) 54.0 ( 1.1)
Nearest-neighbor dist 62.1 ( 1.2) 71.5 ( 0.8) 56.5 ( 1.6)
Principal comp. 68.7 ( 0.1) 77.4 ( 0.3) 0.0 ( 0.0)
Self-Organ. Map 61.6 ( 0.6) 76.3 ( 0.8) 50.7 ( 1.4)
Auto-enc network 66.2 ( 2.7) 79.9 ( 0.2) 56.8 ( 1.4)
Spanning Tree 66.5 ( 2.7) 77.2 ( 0.5) 50.9 ( 1.1)
L_1-ball 55.7 ( 2.4) 48.1 ( 1.3) 56.4 ( 0.1)
k-center 54.9 ( 5.0) 72.9 ( 1.7) 56.4 ( 2.5)
Support vector DD 0.0 ( 0.0) 67.5 ( 3.6) 48.3 ( 1.2)
Minimax Prob. DD 65.2 ( 0.9) 80.0 ( 0.3) 57.3 ( 1.0)
LinProg DD 49.0 ( 3.2) 77.9 ( 0.4) 52.1 ( 2.9)

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