Selecting Optimal Feature Set in High-Dimensional Data by Swarm Search

Algorithm 1

Pseudocode for the WSA algorithm.

Objective function f(x),  x  =  (𝚡_𝟷,𝚡_𝟸,…,𝚡_𝚍)^T

Initialize the population of wolves, x_i(i  =  1,  2,  …,  W)

Define and initialize parameters:

r  =  radius of the visual range

s  =  step size by which a wolf moves at a time

α  =  velocity factor of wolf

p_a  =  a user-defined threshold [0..1],  determines how frequently an enemy appears

WHILE (t  < generations && stopping criteria not met)

 FOR i  =  1 : W // for each wolf

  Prey_new_food_initiatively();

  Generate_new_location();

  // check whether the next location suggested by the random number generator is new. If

not, repeat generating random location.

  IF  (dist(𝚡_𝚒,  𝚡_𝚓) <  r  && x_j  is better as f(𝚡_𝚒) <   f(𝚡_𝚓))

   x_i  moves towards x_j  // x_j  is a better than x_i

  ELSE IF

   x_i  =  Prey_new_food_passively();

  END IF

  Generate_new_location();

  IF  (rand ()>  pa)

   x_i  =  x_i  +  rand ()  +   v; // escape to a new pos.

  END IF

 END FOR

END WHILE