This PR adds the following features to the FeatureComputer:
max_to_min (rate between the minimum and maximum)
count_abs_big (counts big absolute values)
count_abs_ext_big (counts extremely big absolute values)
abs_trend (uses linear regression to find the trend)
mad (mean absolute deviation)
skew (measuring the skewness)
abs_skew (measuring the skewness of the absolute values)
kurtosis (measuring the tailedness of the values)
abs_kurtosis (measuring the tailedness of the absolute values)
hilbert (the mean of the absolute values in hilbert transformed space)
hann (the mean in hann window)
STALTA (the short time average over long time average)
exp_mov_ave (the exponential moving average)
STALTA and exp_mov_ave are calculated for multiple time windows. Note that this time windows needs to be shorter than the total array. We are also calculating these features for subsequences, which are relatively short arrays. Therefore, some changes in the code were necessary to take the correct time window for each array.
This PR adds the following features to the FeatureComputer:
STALTA and exp_mov_ave are calculated for multiple time windows. Note that this time windows needs to be shorter than the total array. We are also calculating these features for subsequences, which are relatively short arrays. Therefore, some changes in the code were necessary to take the correct time window for each array.