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

Table 6

Feature ranking of statistical features for dataset B.

Feature	Feature description	Chi-squared statistics	Info gain	Gain ratio	Average rank score
T2	Root mean square (RMS)	3	4	4	3.67
T18	Geometric mean	4	6	2	4.00
T20	Mean absolute deviation	1	3	8	4.00
T30	Normal negative log likelihood	2	5	9	5.33
T22	Zero crossing rate	15	2	1	6.00
T27	Hjorth parameter—mobility	16	1	3	6.67
T17	Kurtosis factor	5	13	5	7.67
T19	Root-sum-of-squares (RSSQ)	6	14	6	8.67
T9	75th percentile	7	15	7	9.67
T23	Entropy	8	12	11	10.33
T26	Hjorth parameter—activity	10	7	16	11.00
T29	Weibull negative log likelihood	11	8	15	11.33
T6	Standard deviation	12	9	14	11.67
T21	Median absolute deviation	13	10	12	11.67
T8	25th percentile	9	16	10	11.67
T3	Root	14	11	13	12.67
T24	Histogram upper bound	17	17	19	17.67
T4	Max	18	18	18	18.00
T5	Peak-to-peak	19	19	17	18.33
T25	Histogram lower bound	20	20	20	20.00
T28	Hjorth parameter—complexity	22	21	22	21.67
T11	Kurtosis	23	22	21	22.00
T16	Skewness factor	21	23	24	22.67
T15	Clearance factor	24	24	23	23.67
T12	Crest factor	25	25	25	25.00
T7	Median	26	26	27	26.33
T1	Mean	27	27	26	26.67
T10	Skewness	28	28	28	28.00
T14	Impulse factor	29	29	30	29.33
T13	Shape factor	30	30	29	29.67