Dataset Vehicle bus

Basic characteristics Vehicle bus

218

target objects

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

628

outlier objects

18

features

Unsupervised PCA Vehicle bus

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 bus

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

Results Vehicle bus

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

614, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 92.8 ( 0.1) 97.2 ( 0.1) 59.9 ( 0.3)
Min.Cov.Determinant 93.2 ( 0.2) 93.2 ( 0.2) 64.4 ( 0.2)
Mixture of Gaussians 83.1 ( 0.3) 97.0 ( 0.3) 63.3 ( 0.3)
Naive Parzen 85.8 ( 0.4) 85.8 ( 0.4) 77.1 ( 0.2)
Parzen 90.6 ( 0.6) 92.6 ( 2.6) 77.1 ( 0.2)
k-means 76.2 ( 0.9) 90.4 ( 1.0) 70.6 ( 0.8)
1-Nearest Neighbors 56.6 ( 5.6) 95.1 ( 0.2) 76.5 ( 0.5)
k-Nearest Neighbors 56.6 ( 5.6) 95.1 ( 0.2) 76.5 ( 0.5)
knn, opt-AUC 64.5 ( 4.5) 95.0 ( 0.3) 76.3 ( 0.4)
Nearest-neighbor dist 84.1 ( 3.1) 93.1 ( 0.8) 63.0 ( 2.2)
Principal comp. 59.9 ( 0.2) 92.2 ( 0.5) 0.0 ( 0.0)
Self-Organ. Map 64.3 ( 1.1) 92.8 ( 0.4) 53.6 ( 1.2)
Auto-enc network 69.5 ( 4.6) 93.5 ( 0.4) 71.5 ( 1.4)
Spanning Tree 87.6 ( 4.0) 95.7 ( 0.2) 50.6 ( 2.4)
L_1-ball 62.1 ( 1.6) 41.7 ( 0.9) 59.9 ( 0.3)
k-center 64.7 ( 5.1) 89.7 ( 0.6) 71.9 ( 1.2)
Support vector DD 0.2 ( 0.2) 91.8 ( 2.3) 74.2 ( 0.7)
Minimax Prob. DD 86.4 ( 0.9) 95.6 ( 0.1) 76.7 ( 0.3)
LinProg DD 45.6 ( 6.0) 92.3 ( 0.3) 49.1 ( 3.9)

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