Algorithm 1 Model selection for Kernel_SVDD

Input: Training dataset Train = {x1, x2, …, xn}

Support vector SVS of kernel_SVDD

Process:

1: while (1) do

2: Sample T(1)~N(0, ID/δ2);

3: Apply the Toeplitz transformation of T(1) to form a D-dimensional feature matrix TD;

4: Train the training set Train to obtain decision model TRFF_f using TRFF algorithm;

5: Calculate the over-fitting error: error_over

6: if error_over = 0

7:   Calculate the under-fitting error error_under;

8:   if error_under < error_underτ

9:    break;

10:   else

11:    continue;

12:   end if;

13: else

14:  continue;

15: end if;

16: end while;

Output: Random feature matrix of optimal model TD