Skip to main content

Table 2 On-board runtimes during feature calculations. Where features have been grouped in one row the total runtime for the calculations of all features is total listed. Under “Note” any dependencies for the calculation of the feature are listed. “Gross time” identifies the total runtime for the listed feature and its dependencies

From: An evaluation of machine learning classifiers for next-generation, continuous-ethogram smart trackers

Feature(abbreviation) Net time(ms) Gross time(ms) Number of features calculated Note
Mean(mean) 0.021 0.021 3  
Variance(var) 0.025 0.046 3 mean
Standard deviation(sd) 0.003 0.049 3 mean, var
Coefficient of variance 0.002 0.051 3 mean, sd
Skewness 0.361 0.41 3 mean, sd
Kurtosis 0.367 0.416 3 mean, sd
Maximum(max) 0.029 0.029 3  
Minimum(min) 0.029 0.029 3  
Range 0.001 0.059 3 max, min
Euclidean norm 0.017 0.017 3  
Covariance(cov) 0.033 0.054 3 mean
Correlation 0.002 0.084 3 sd, cov
Mean difference(meandiff) 0.026 0.026 3  
Standard deviation of difference 0.035 0.061 3 meandiff
Variance of static body acceleration
Variance of dynamic body acceleration
Mean dynamic body acceleration
Maximum dynamic body acceleration
Overall dynamic body acceleration
0.279 0.279 13  
Pitch 0.039 0.06 1 mean
Roll 0.049 0.07 1 mean
Mean difference of continuous points(meandl) 0.178 0.178 3  
Variance of difference of continuous points 0.025 0.203 3 meandl
Main frequency
Amplitude of main frequency
0.258 0.258 6  
25% quantile
50% quantile
75% quantile
0.94 0.94 9  
\