Improved Instance Selection Methods for Support Vector Machine Speed Optimization

Algorithm 3

FFA_instance_selection.

Notation

= number of fireflies

NS = number of selected instances

= Max Generation

Min = Minimum number of selected instances

= Initial Attractiveness Value

= alpha

= Gamma

= Objective function, where and = number of fireflies

IM = Instance Mask or Subset of instances.

= Size of Instance Mask

= Dataset

GB = Global Best

TS = Training subset

= light intensity

Input: , , , ,  , ,

Output: TS

(1) Define

(2) Initialize IM of each firefly

(3) Evalute to determine for each firefly

(4) Select firefly with the highest , and save in GB

(5) while ()

  (5.1) for   to

     (5.1.1) for   to

       (5.1.1.1) If ()

         (5.1.1.1.1) Move firefly towards firefly

       (5.1.1.2) end if

       (5.1.1.3) Calculate attrativeness variance with distance using

       (5.1.1.4) Evaluate to determine new fitness value for firefly

       (5.1.1.5) Update light intensity of firefly

     (5.1.2) end  

  (5.2) end  

  (5.3) Update GB

(6) end while

(7) Calculate NS of GB

(8) if NS in GB < Min

  (8.1) update GB by assigning 1 to (Min – NS) instances that was not selected

(9) End if

(10) For   to

  (10.1) If   in GB is equal to 1

    (10.1.1)

  (10.2) End if

(11) Output TS