Dataset Delft pump 5x1 noisy

Basic characteristics Delft pump 5x1 noisy

78

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. All loads with one speed situations are used in the target class. Download mat-file with Prtools dataset.

222

outlier objects

64

features

Unsupervised PCA Delft pump 5x1 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 5x1 noisy

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

Results Delft pump 5x1 noisy

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

548, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 96.3 (0.0) 94.7 (0.0) 48.1 (0.0)
Min.Cov.Determinant NaN (0.0) NaN (0.0) 58.8 (0.0)
Mixture of Gaussians 94.9 (0.5) 90.9 (1.8) 67.7 (2.4)
Naive Parzen 73.6 (0.0) 73.6 (0.0) 63.9 (0.0)
Parzen 93.0 (0.0) 87.4 (0.0) 86.3 (0.0)
k-means 75.0 (4.2) 77.1 (2.1) 45.1 (1.6)
1-Nearest Neighbors 93.0 (0.0) 87.5 (0.0) 86.4 (0.0)
k-Nearest Neighbors 93.0 (0.0) 87.5 (0.0) 86.4 (0.0)
Nearest-neighbor dist 96.9 (0.0) 96.3 (0.0) 89.0 (0.0)
Principal comp. 94.5 (0.0) 93.1 (0.0) 54.9 (0.0)
Self-Organ. Map 90.9 (2.8) 85.5 (1.0) 80.7 (4.8)
Auto-enc network 72.0 (13.0) 77.6 (5.5) 47.4 (0.0)
MST 94.1 (0.0) 89.6 (0.0) 87.7 (0.0)
L_1-ball 53.4 (0.0) 61.8 (0.0) 60.8 (0.0)
k-center 83.4 (1.5) 75.2 (3.1) 72.5 (4.5)
Support vector DD 92.9 (0.4) 70.3 (0.2) 87.1 (1.5)
Minimax Prob. DD 92.6 (0.0) 87.4 (0.0) 87.1 (0.0)
LinProg DD 90.1 (0.0) 79.4 (0.0) 84.7 (0.0)
Lof DD 85.5 (0.0) 90.4 (0.0) 75.2 (0.0)
Lof range DD 88.0 (0.0) 89.5 (0.0) 71.5 (0.0)
Loci DD 86.0 (0.0) 77.8 (0.0) 63.0 (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