Sliding Measures

Sliding measures [1]_ define the distance between time series \(X\) and time series \(Y\) by finding the largest correlation between \(X\) and all shifted versions of \(Y\) , where each shifted version is created by moving all entries in \(Y\) towards right by \(s\) positions. In this process, we create a cross-correlation sequence, \(CC_{w}(\vec{x}, \vec{y})\) with \(w\in{1, 2, ..., 2m-1}\) of length \((2m-1)\) that contains the inner product of two time series in every possible shift.

Normalized Cross-Correlation

tsdistance.sliding.NCC(x, y)

Return the full cross-correlation sequence.

Biased Normalized Cross-Correlation

tsdistance.sliding.NCCb(x, y)

Return the biased normalized cross-correlation sequence.

Unbiased Normalized Cross-Correlation

tsdistance.sliding.NCCu(x, y)

Return the unbiased normalized cross-correlation sequence.

Coefficient Normalized Cross-Correlation

tsdistance.sliding.NCCc(x, y)

Return the coefficient normalized cross-correlation sequence.

Reference .. [1] John Paparrizos et al. “Debunking Four Long-Standing Misconceptions ofTime-Series Distance Measures”. In:ACM SIGMOD(2020)