repr_featrend: FeaTrend representation of time series
Description
The repr_featrend computes representation of time series based on feature extraction from bit-level (trending) representation.
Usage
repr_featrend(x, func, pieces = 2L, order = 4L)
Arguments
x
the numeric vector (time series)
func
the function of aggregation, can be sumC or maxC or similar aggregation function
pieces
the number of parts of time series to split (default to 2)
order
the order of simple moving average (default to 4)
Value
the numeric vector of the length pieces
Details
FeaTrend is method of time series representation based on feature extraction from run lengths (RLE) of bit-level (trending) representation.
It extracts number of features from trending representation based on number of pieces defined.
From every piece, 2 features are extracted. You can define what feature will be extracted,
recommended functions are max and sum. For example if max is selected, then maximum value of run lengths of ones and zeros are extracted.
# NOT RUN {# default settingsrepr_featrend(rnorm(50), maxC)
# compute FeaTrend for 4 pieces and make more smoothed ts by order = 8repr_featrend(rnorm(50), sumC, 4, 8)
# }