Algorithm 1. Auto-Labeling Data for CTL

Input: Original Time series data$X={[x_1,x_2,x_3,\mathrm{...},x_N]}^T$,$\omega$>0, which represents the proportion threshold parameter of the trend definition

Output: The label vector $Y={[{label}_1,{label}_2,{label}_3,\mathrm{...},{label}_N]}^T$

Initialization of related variables:

FP=x1, which represents the first price obtained by the algorithm; xH =x1, used to mark the highest price; HT=t1, used to mark the time when the highest price occurs; xL =x1, used to mark the lowest price; LT=t1, used to mark the time when the lowest price occurs; Cid=0, used to mark the current direction of labeling; FP_N =0, the index of the highest or lowest point obtained initially.

fori = 1:N   if (xi > FP + x1*$\omega$)     Set [xH, HT, FP_N, Cid ] = [ xi, ti, i,1] and end for    if (xi < FP- x1*$\omega$)     Set [xL, LT, FP_N, Cid ] = [ xi, ti, i, −1] and end for end for i for i = FP_N+1:N    if (Cid > 0)     if (xi > xH)      Set [xH, HT] = [ xi, ti ]     if (xi < xH - xH*$\omega$ and LT <= HT)      for j = 1:N       if (tj > LT and tj <= HT)        Set yj = 1      end for j       Set [xL, LT, Cid] = [ xi, ti, −1]     if (Cid < 0)      if (xi < xL)       Set [xL, LT] = [ xi, ti ]      if (xi > xL + xL *$\omega$ and HT <= LT)      for j = 1:N       if (tj > HT and tj < = LT)        Set yj = −1      end for j      Set[xH, HT, Cid] = [ xi, ti, 1] end for i