Dataset Satellite, damp grey soil

Basic characteristics Satellite, damp grey soil

415

target objects

The Landsat Satellite database from the Statlog Project. The classification associated with the central pixel in each 3x3 neighbourhood in a satellite image, based on multi-spectral values of pixels. Class damp grey soil is used as target class. Download mat-file with Prtools dataset.

4020

outlier objects

36

features

Unsupervised PCA Satellite, damp grey soil

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 Satellite, damp grey soil

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

Results Satellite, damp grey soil

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

624, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 83.0 ( 0.0) 83.0 ( 0.0) 87.0 ( 0.0)
Min.Cov.Determinant 78.6 ( 0.1) 78.6 ( 0.1) 81.4 ( 0.0)
Mixture of Gaussians 67.3 (23.3) 75.5 (20.4) 86.6 ( 0.3)
Naive Parzen 93.6 ( 0.1) 93.5 ( 0.1) 86.4 ( 0.0)
Parzen 42.9 ( 0.0) 39.9 ( 0.0) 43.2 ( 0.0)
k-means 79.4 (19.4) 88.7 ( 0.4) 88.3 ( 0.9)
1-Nearest Neighbors 38.4 ( 0.0) 90.6 ( 0.0) 90.3 ( 0.0)
k-Nearest Neighbors 38.4 ( 0.0) 90.6 ( 0.0) 90.3 ( 0.0)
knn, opt-AUC 45.2 ( 0.7) 90.6 ( 0.0) 90.1 ( 0.3)
Nearest-neighbor dist 41.6 ( 1.2) 88.2 ( 0.0) 84.5 ( 0.0)
Principal comp. 79.4 ( 0.0) 79.8 ( 0.0) 78.3 ( 0.0)
Self-Organ. Map 90.1 ( 0.5) 89.1 ( 0.4) 90.4 ( 0.3)
Auto-enc network 79.1 ( 1.9) 78.1 ( 0.1) 78.2 ( 0.0)
Spanning Tree 40.6 ( 0.2) 90.6 ( 0.0) 90.5 ( 0.0)
L_1-ball 90.8 ( 0.0) 91.2 ( 0.0) 75.7 ( 0.0)
k-center 46.8 ( 2.1) 85.0 ( 1.5) 87.7 ( 2.4)
Support vector DD 0.0 ( 0.0) 21.1 ( 0.0) 0.0 ( 0.0)
Minimax Prob. DD 81.7 ( 0.0) 90.8 ( 0.0) 87.0 ( 0.0)
LinProg DD 19.5 ( 1.5) 51.8 (22.4) 20.5 ( 1.1)

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