Dataset Delft pump 3x2 noisy

Basic characteristics Delft pump 3x2 noisy

83

target objects

The noisy Delft pump dataset, with some description in A. Ypma, Learning methods for machine vibration analysis and health monitoring, thesis Delft university of Technology, 2001. From a submersible pump 5 vibration measurements are taken under different normal and abnormal conditions. From the time signals an 64D envelope spectrum is derived. The 5 measurements are used as independent objects. Some loads with some speeds situations are used in the target class. Download mat-file with Prtools dataset.

277

outlier objects

64

features

Unsupervised PCA Delft pump 3x2 noisy

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 Delft pump 3x2 noisy

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

Results Delft pump 3x2 noisy

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

549, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 90.6 (0.0) 87.3 (0.0) 49.7 (0.0)
Min.Cov.Determinant NaN (0.0) NaN (0.0) 60.2 (0.0)
Mixture of Gaussians 86.4 (1.4) 78.5 (1.8) 64.3 (2.5)
Naive Parzen 67.0 (0.0) 67.0 (0.0) 58.2 (0.0)
Parzen 79.0 (0.0) 70.6 (0.0) 68.0 (0.0)
k-means 71.7 (2.9) 64.7 (2.0) 48.8 (3.0)
1-Nearest Neighbors 79.0 (0.0) 70.7 (0.0) 68.3 (0.0)
k-Nearest Neighbors 79.0 (0.0) 70.7 (0.0) 68.3 (0.0)
Nearest-neighbor dist 76.9 (0.0) 79.5 (0.0) 66.8 (0.0)
Principal comp. 86.3 (0.0) 80.9 (0.0) 50.9 (0.0)
Self-Organ. Map 84.1 (0.4) 69.6 (0.9) 68.8 (0.8)
Auto-enc network 75.4 (5.5) 66.0 (2.1) 51.9 (0.0)
MST 83.7 (0.0) 71.9 (0.0) 72.1 (0.0)
L_1-ball 52.0 (0.0) 62.8 (0.0) 60.6 (0.0)
k-center 78.3 (0.4) 68.6 (1.3) 67.0 (0.8)
Support vector DD 78.0 (0.3) 60.7 (0.1) 67.8 (0.8)
Minimax Prob. DD 78.7 (0.0) 70.7 (0.0) 68.3 (0.0)
LinProg DD 81.1 (0.0) 64.7 (0.0) 68.6 (0.0)
Lof DD 77.9 (0.0) 78.8 (0.0) 65.7 (0.0)
Lof range DD 76.0 (0.0) 75.8 (0.0) 61.3 (0.0)
Loci DD 78.1 (0.0) 62.4 (0.0) 54.8 (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