Algorithm 1. Data Pre-process

Input: A time-series matrix $d$, sliding window size $l$, sliding window step $s$

Output: A matrix of final train samples, A matrix of final test samples

1. result = []

2. Initialize the flag number start = 0

3. while start < len($d$)

4.  if start + $l$ < len($d$)

5.   temp = $d$.iloc[start:start+$l$, :]            //data segmentation

6.   reshaped_temp = reshape(temp, [1, $l\times$$d$.shape [1]]) //data reshaping

7.   result.append(reshaped_temp)

8.    start = start +$s$

9.  end if

10. end while

11. X_train, X_test = train_test_split(test_size = 0.3)

12. //data standardization

13. for col in X_train.cols

14.  col_max = max(X_train.iloc[:, col])

15.  col_min = min(X_train.iloc[:, col])

16.  X_train.iloc[:, col] = (X_train.iloc[:, col] – col_min)/(col_max – col_min)

17.  X_test.iloc[:, col] = (X_test.iloc[:, col] – col_min)/(col_max – col_min)

18. end for

19. return X_train, X_test