# tsfuse.transformers

## tsfuse.transformers.boolean

class tsfuse.transformers.boolean.Greater(*parents, **kwargs)

Element-wise greater than comparison

transform(x, y, **kwargs)

Compute $$x > y$$

class tsfuse.transformers.boolean.Less(*parents, **kwargs)

Element-wise less than comparison

transform(x, y, **kwargs)

Compute $$x < y$$

class tsfuse.transformers.boolean.Equal(*parents, **kwargs)

Element-wise equality comparison

transform(x, y, **kwargs)

Compute $$x = y$$

class tsfuse.transformers.boolean.NotEqual(*parents, **kwargs)

Element-wise inequality comparison

transform(x, y, **kwargs)

Compute $$x \neq y$$

## tsfuse.transformers.frequency

class tsfuse.transformers.frequency.FFT(*parents, attr='abs', axis='time', **kwargs)

Fast Fourier transform

Parameters
• attr ({'real', 'imag', 'abs', 'angle'}, optional) – Return the real part (‘real’), imginary part (‘imag’), absolute value (‘abs’), or angle in degrees (‘angle’). Default: ‘abs’

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: ‘time’

transform(x, **kwargs)

Compute the fast Fourier transform of each time series in x

Parameters

x (Collection) – Time series data.

class tsfuse.transformers.frequency.CWT(*parents, wavelet='ricker', width=1, axis=None, **kwargs)

Continuous wavelet transform

Parameters
• wavelet ({'ricker'}, optional) – Wavelet type. Default: ‘ricker’

• width (int, optional) – Wavelet width. Default: 1

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: ‘time’

transform(x, **kwargs)

Compute continous wavelet transform for each time series in x

Parameters

x (Collection) – Time series data.

class tsfuse.transformers.frequency.PowerSpectralDensity(*parents, axis=None, **kwargs)

Power spectral density

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: ‘time’

transform(x, **kwargs)

Compute power spectral density using Welch’s method, for each time series in x

Parameters

x (Collection) – Time series data.

## tsfuse.transformers.geometry

class tsfuse.transformers.geometry.Norm(*parents, p=2, **kwargs)

Vector norm

Parameters

p (int, optional) – Order of the vector norm. Default: 2

transform(x, **kwargs)

Compute the vector norm of order $$p$$ over the dimensions of x

Parameters

x (Collection) – Multivariate time series data with least 2 dimensions.

class tsfuse.transformers.geometry.Resultant(*parents, **kwargs)

Euclidean norm

transform(x, **kwargs)

Compute the Euclidean norm over the dimensions of x

Parameters

x (Collection) – Multivariate time series data with least 2 dimensions.

class tsfuse.transformers.geometry.Angle(*parents, **kwargs)

Angle defined by three points

transform(p1, p2, p3, **kwargs)

Compute the angle $$\mathbf{\theta}$$ defined by three points p1, p2, p3 as shown in the figure below:

This function uses the following formula for computing $$\theta$$:

$$\theta = \mathrm{arccos}\Bigg( \frac{\overrightarrow{p1p2}~\cdot~\overrightarrow{p2p3}}{||\overrightarrow{p1p2}||~||\overrightarrow{p2p3}||} \Bigg)$$

Parameters
• p1 (Collection) – 2D/3D coordinates of point 1.

• p2 (Collection) – 2D/3D coordinates of point 2.

• p3 (Collection) – 2D/3D coordinates of point 3.

## tsfuse.transformers.mathematics

transform(x, y, **kwargs)

Compute $$x + y$$

class tsfuse.transformers.mathematics.Subtract(*parents, **kwargs)

Element-wise subtraction

transform(x, y, **kwargs)

Compute $$x - y$$

class tsfuse.transformers.mathematics.Multiply(*parents, **kwargs)

Element-wise multiplication

transform(x, y, **kwargs)

Compute $$x \cdot y$$

class tsfuse.transformers.mathematics.Divide(*parents, **kwargs)

Element-wise division

transform(x, y, **kwargs)

Compute $$x / y$$

class tsfuse.transformers.mathematics.Negative(*parents, **kwargs)

Element-wise negation

transform(x, **kwargs)

Compute $$-x$$

class tsfuse.transformers.mathematics.Reciprocal(*parents, **kwargs)

Element-wise multiplicative inverse

transform(x, **kwargs)

Compute $$1/x$$

class tsfuse.transformers.mathematics.Square(*parents, **kwargs)

Element-wise square

transform(x, **kwargs)

Compute $$x^2$$

class tsfuse.transformers.mathematics.Exponent(*parents, a=2, **kwargs)

Element-wise exponent.

Parameters

a (int, optional) – Exponent. Default: 2

transform(x, **kwargs)

Compute $$x^a$$

class tsfuse.transformers.mathematics.Sqrt(*parents, **kwargs)

Element-wise square root

transform(x, **kwargs)

Compute $$\sqrt{x}$$

class tsfuse.transformers.mathematics.Abs(*parents, **kwargs)

Element-wise absolute value

transform(x, **kwargs)

Compute $$|x|$$

class tsfuse.transformers.mathematics.Sum(*parents, axis=None, **kwargs)

Summation

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

Compute the sum along the given axis

class tsfuse.transformers.mathematics.CumSum(*parents, axis=None, **kwargs)

Cumulative summation

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

Compute cumulative sum along the given axis

class tsfuse.transformers.mathematics.Diff(*parents, axis=None, **kwargs)

First-order derivative

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

Compute difference $$v_{i+1} - v_i$$ for all pairs of consecutive value $$v_i$$ and $$v_{i+1}$$ along the given axis

class tsfuse.transformers.mathematics.Roots(*parents, axis=None, **kwargs)

Roots of a polynomial

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

For the axis to which this transformer is applied, the values $$p_0, p_1, ..., p_n$$ are interpreted as the coefficients of a polynomial of degree n:

$$p_0 \cdot x^n + p_1 \cdot x^{n-1} + ... + p_n$$

For each polynomial, this transformer computes the values where the result of the polynomial equals zero.

Notes

Only the real roots are returned (i.e., no complex roots)

class tsfuse.transformers.mathematics.Average(*parents, **kwargs)

Element-wise average

transform(x, y, **kwargs)

Compute $$\frac{x+y}{2}$$

class tsfuse.transformers.mathematics.Difference(*parents, rel=False, **kwargs)

Element-wise difference

Parameters

rel (bool, optional) – Compute the relative difference. Default: False

transform(x, y, **kwargs)

Compute $$|x-y|$$ and divide by $$|x|$$ if rel is true

class tsfuse.transformers.mathematics.Ratio(*parents, **kwargs)

Element-wise ratio

transform(x, y, **kwargs)

Compute $$x/y$$

## tsfuse.transformers.peaks

class tsfuse.transformers.peaks.NumberPeaks(*parents, support=1, axis=None, **kwargs)

Number of peaks

Parameters
• support (int, optional) – Minimum support of each peak. Default: 1

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

For each series in x, compute the number of peaks that have a support larger than the given minimum support. The support of a peak is defined as the length of the largest subsequence around the peak where the peak has the largest value.

class tsfuse.transformers.peaks.NumberPeaksCWT(*parents, max_width=1, axis=None, **kwargs)

Number of peaks estimated using a continous wavelet transform

Parameters
• max_width (int, optional) – Maximum width of the wavelet. Default: 1

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

For each series in x, estimate the number of peaks using scipy.signal.find_peaks_cwt where widths = [1, ..., max_width]

## tsfuse.transformers.sampling

class tsfuse.transformers.sampling.Resample(*parents, num=None, axis=None, **kwargs)
Parameters
• num (int) – New number of samples.

• axis ({'time', 'dims'}, optional) – Time direction: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

Resample x in the time direction to num samples using scipy.signal.resample

## tsfuse.transformers.statistics

class tsfuse.transformers.statistics.Length(*parents, axis=None, **kwargs)

Number of samples

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Sum(*parents, axis=None, **kwargs)

Summation

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

transform(x, **kwargs)

Compute the sum along the given axis

class tsfuse.transformers.statistics.Mean(*parents, axis=None, **kwargs)

Arithmetic mean

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Median(*parents, axis=None, **kwargs)

Middle value of the sorted values

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Min(*parents, axis=None, **kwargs)

Smallest value

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.ArgMin(*parents, first=True, rel=False, axis=None, **kwargs)

Location of smallest value

Parameters
• first (bool, default: True) – If true, return first location. Otherwise, return last location.

• rel (bool, default: False) – Return location relative to total length (i.e., as a number in [0, 1])

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Max(*parents, axis=None, **kwargs)

Largest value

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.ArgMax(*parents, first=True, rel=False, axis=None, **kwargs)

Location of largest value

Parameters
• first (bool, default: True) – If true, return first location. Otherwise, return last location.

• rel (bool, default: False) – Return location relative to total length (i.e., as a number in [0, 1])

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Variance(*parents, axis=None, **kwargs)
Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.StandardDeviation(*parents, axis=None, **kwargs)

Standard deviation

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Skewness(*parents, axis=None, **kwargs)
Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Kurtosis(*parents, axis=None, **kwargs)
Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SpectralMoment(*parents, r=1, origin=False, axis=None, **kwargs)

Spectral moment

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SpectralMean(*parents, origin=False, axis=None, **kwargs)

Spectral mean

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SpectralVariance(*parents, origin=False, axis=None, **kwargs)

Spectral variance

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SpectralSkewness(*parents, origin=False, axis=None, **kwargs)

Spectral skewness

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SpectralKurtosis(*parents, origin=False, axis=None, **kwargs)

Spectral kurtosis

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Quantile(*parents, q=0.5, axis=None, **kwargs)

q-Quantile

Parameters
• q (float, default: 0.5) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.IndexMassQuantile(*parents, q=0.5, rel=False, axis=None, **kwargs)

Index mass q-quantile

Parameters
• q (float, default: 0.5) –

• rel (bool, default: False) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Energy(*parents, axis=None, **kwargs)

Sum of the squared values

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.EnergyRatio(*parents, chunks=10, axis=None, **kwargs)

Energy ratio

Divides the values into bins and computes the sum of squares of each bin divided by the total sum of squares of all bins.

Parameters
• chunks (int) – Number of bins.

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Entropy(*parents, axis=None, **kwargs)
Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SampleEntropy(*parents, axis=None, **kwargs)

Sample entropy

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.BinnedDistribution(*parents, bins=10, axis=None, **kwargs)

Binned distribution

Divides the values into bins and computes the fraction of values in each bin.

Parameters
• bins (int, default: 10) – Number of bins.

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.BinnedEntropy(*parents, bins=10, axis=None, **kwargs)

Binned entropy

Divides the values into bins and computes the entropy of the values in each bin.

Parameters
• bins (int, default: 10) – Number of bins.

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.C3(*parents, lag=1, axis=None, **kwargs)

Non-linearity

Parameters
• lag (int, default: 1) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.CID(*parents, axis=None, **kwargs)

Complexity estimate

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.CountAboveMean(*parents, axis=None, **kwargs)

Number of values greater than the mean

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.CountBelowMean(*parents, axis=None, **kwargs)

Number of values smaller than the mean

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.RangeCount(*parents, min=- 1, max=1, axis=None, **kwargs)

Number of values in range

Parameters
• min (int, default: -1) – Lower bound of the range.

• max (int, default: 1) – Upper bound of the range.

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.ValueCount(*parents, value=0, axis=None, **kwargs)

Number of occurrences of a value

Parameters
• value (int, default: 0) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.Outliers(*parents, r=3, rel=False, axis=None, **kwargs)

Number of outliers, based on number of unit standard deviations from mean

Parameters
• r (int, default: 3) – Number of unit standard deviations from mean.

• rel (bool, default: False) – Return the number of outliers relative to the total number of values.

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.AutoCorrelation(*parents, axis=None, **kwargs)

Auto-correlation coefficients

Returns the auto-correlation coefficients for lags [1, ..., t] where t is the total number of values.

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.CrossCorrelation(*parents, axis=None, **kwargs)

Cross-correlation coefficients of two series

Returns the cross-correlation coefficients for lags [1, ..., t] where t is the total number of values in each series.

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.AutoRegressiveCoefficients(*parents, axis=None, **kwargs)

Auto regressive coefficients

Fits an autoregressive (AR) model and returns the coefficients of the model.

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.HighVariance(*parents, threshold=1, axis=None, **kwargs)

Check if the series has a high variance

Returns true if the variance is larger than a given threshold.

Parameters
• threshold (float, default: 1) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.HighStandardDeviation(*parents, r=1, axis=None, **kwargs)

Check if the series has a high standard deviation

Returns true if

$StandardDeviation(x) > r * (Max(x) - Min(x))$
Parameters
• r (float, default: 1) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SymmetryLooking(*parents, r=1, axis=None, **kwargs)

Check whether the series looks symmetrical

Returns true if

$|Mean(x) - Median(x)| < r * (Max(x) - Min(x))$
Parameters
• r (float, default: 1) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.NumberCrossings(*parents, threshold=0, axis=None, **kwargs)

Number of times that the series crosses a given threshold

Parameters
• threshold (float, default: 0) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.LinearTrend(*parents, axis=None, **kwargs)

Linear trend statistics

Using scipy.stats.linregress this transformer computes the following linear trend statistics:

• Slope

• Intercept

• Pearson correlation coefficient

• p-value of the hypothesis test where the null hypothesis is that the slope is zero

• Standard error of the slope

• Standard error of the intercept

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.LongestStrikeAboveMean(*parents, axis=None, **kwargs)

Largest number of consecutive values greather than the mean

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.LongestStrikeBelowMean(*parents, axis=None, **kwargs)

Largest number of consecutive values smaller than the mean

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.SumChange(*parents, abs=False, axis=None, **kwargs)

Sum of differences between consecutive timestamps

Parameters
• abs (bool, default: False) – Take the absolute value of the differences

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.MeanChange(*parents, abs=False, axis=None, **kwargs)

Average of differences between consecutive timestamps

Parameters
• abs (bool, default: False) – Take the absolute value of the differences

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.MeanSecondDerivativeCentral(*parents, axis=None, **kwargs)

Mean of the central approximation of the second derivative

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.TimeReversalAsymmetryStatistic(*parents, lag=1, axis=None, **kwargs)

Time reversal asymmetry statistic

Parameters
• lag (int, default: 1) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.FriedrichCoefficients(*parents, m=1, r=10, axis=None, **kwargs)

Coefficients of a polynomial fitted to the deterministic dynamics of a Langevin model.

Parameters
• m (int, default: 1) –

• r (int, default: 10) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.statistics.MaxLangevinFixedPoint(*parents, m=1, r=10, axis=None, **kwargs)

Largest fixed point of a polynomial fitted to the deterministic dynamics of a Langevin model.

Parameters
• m (int, default: 1) –

• r (int, default: 10) –

• axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

## tsfuse.transformers.uniqueness

class tsfuse.transformers.uniqueness.HasDuplicate(*parents, axis=None, **kwargs)

Check whether the series has a duplicated value

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.uniqueness.HasDuplicateMin(*parents, axis=None, **kwargs)

Check whether the series has a duplicated minimum

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.uniqueness.HasDuplicateMax(*parents, axis=None, **kwargs)

Check whether the series has a duplicated maximum

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.uniqueness.NumberUniqueValues(*parents, rel=True, axis=None, **kwargs)

Number of unique values in the series

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.

class tsfuse.transformers.uniqueness.SumReoccurringValues(*parents, axis=None, **kwargs)

Sum of values that occur at least twice

Parameters

axis ({'time', 'dims'}, optional) – Direction of time: timestamps (‘time’) or dimensions (‘dims’). Default: first axis with more than one value.