Dataset Vehicle opel

Basic characteristics Vehicle opel

212

target objects

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

634

outlier objects

18

features

Unsupervised PCA Vehicle opel

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 opel

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

Results Vehicle opel

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

615, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 71.6 ( 0.4) 83.1 ( 0.3) 54.7 ( 0.4)
Min.Cov.Determinant 83.1 ( 0.2) 83.1 ( 0.2) 63.4 ( 0.8)
Mixture of Gaussians 67.2 ( 0.8) 74.9 ( 3.0) 54.5 ( 0.5)
Naive Parzen 67.3 ( 0.6) 67.3 ( 0.6) 52.4 ( 0.8)
Parzen 62.1 ( 3.0) 72.7 ( 1.8) 52.4 ( 0.8)
k-means 57.3 ( 2.1) 69.8 ( 1.5) 50.9 ( 2.5)
1-Nearest Neighbors 49.6 ( 3.5) 76.9 ( 0.3) 47.8 ( 1.2)
k-Nearest Neighbors 49.6 ( 3.5) 76.9 ( 0.3) 47.8 ( 1.2)
knn, opt-AUC 50.4 ( 2.7) 77.0 ( 0.3) 51.6 ( 0.6)
Nearest-neighbor dist 61.3 ( 1.5) 71.8 ( 0.6) 45.4 ( 1.5)
Principal comp. 59.7 ( 0.4) 69.1 ( 0.4) 0.0 ( 0.0)
Self-Organ. Map 54.8 ( 0.7) 75.0 ( 1.1) 48.5 ( 1.8)
Auto-enc network 60.6 ( 1.7) 75.7 ( 0.4) 52.1 ( 2.7)
Spanning Tree 64.7 ( 0.4) 76.7 ( 0.5) 45.8 ( 5.7)
L_1-ball 56.6 ( 1.0) 47.3 ( 1.0) 54.7 ( 0.4)
k-center 49.9 ( 2.3) 72.1 ( 1.7) 50.0 ( 1.8)
Support vector DD 0.0 ( 0.0) 67.8 ( 2.2) 48.7 ( 0.7)
Minimax Prob. DD 61.8 ( 0.8) 78.1 ( 0.3) 48.5 ( 1.1)
LinProg DD 48.2 ( 1.2) 76.0 ( 0.2) 46.8 ( 2.7)

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