Algorithm 3 Motif selection algorithm with prediction.

Require: B, the Codebook of the motifs; S(i), the segment to be compressed; d(·), a dissimilarity metric; dth, the compression distortion threshold; PM, the prediction matrix; X⌂(i−N), the previous best motif.

dmin ← ∞

$\overline{O}\leftarrow Q((1/N){\sum }_{k=1}^N{S}_{i+k-1})$

G̅ ← Q(GAIN(S(i)))

sort B according to the row of PM related to X̅(i−N)

for all X(k) ∈ B do

XR(k) ← X(k)G̅ + O̅

if d(S(i),XR(k)) < dminthen

X̂(i) = XR (k)

dmin = d(S(i),XR(k))

end if

if dmin ≤ dth then

Break

end if

end for

Return k