Optimal Statistical Feature Subset Selection for Bearing Fault Detection and Severity Estimation

Table 15

Comparison of normal negative log likelihood for Gaussian mixture models with normal negative log likelihood for single Gaussian for selected samples.

Sample no.	Normal negative log likelihood for single Gaussian	Normal negative log likelihood for GMM
1	−143.347	−466.000
2	−567.550	−987.950
3	901.479	901.479

The values shown are taken before min-max scaling was applied.