├── README.md
├── Fitness_func.m
└── ensemble_PSO.m


/README.md:
--------------------------------------------------------------------------------
1 | # Pso-based-weighted-ensemble
2 | This is the Matlab implementation of "Elham Ghazizadeh, Bahareh Nikpour, Hossein Nezamabadi-pour and Dariush Abbasi-Moghadam, ” A PSO-based Weighting Method to Enhance Machine Learning Techniques for Cooperative Spectrum Sensing in CR Networks”, The 1st  Conference on Swarm Intelligence and Evolutionary Computation (CSIEC2016), IEEE, 2016." 
3 | 


--------------------------------------------------------------------------------
/Fitness_func.m:
--------------------------------------------------------------------------------
 1 | function [fitness]=Fitness_func(PredClass_SVM,PredClass_KNN,PredClass_NB,N,dim,testdata,testLabels,population)
 2 | 
 3 | for i=1:N
 4 |     for j=1:size(testLabels,1)
 5 |         
 6 |         if PredClass_SVM(j)==PredClass_KNN(j)
 7 |             if population(i,1)+population(i,2)>population(i,3)
 8 |                 particle_lable(j,1)=PredClass_KNN(j);
 9 |             else 
10 |                 particle_lable(j,1)=PredClass_NB(j);
11 |             end
12 |         elseif PredClass_SVM(j)==PredClass_NB(j)
13 |             if population(i,2)+population(i,3)>population(i,1)
14 |                 particle_lable(j,1)=PredClass_SVM(j);
15 |             else 
16 |                 particle_lable(j,1)=PredClass_KNN(j);
17 |             end
18 |         elseif PredClass_KNN(j)==PredClass_NB(j)
19 |             if population(i,1)+population(i,3)>population(i,2)
20 |                 particle_lable(j,1)=PredClass_KNN(j);
21 |             else 
22 |                 particle_lable(j,1)=PredClass_SVM(j);
23 |             end
24 | %         else
25 | %             particle_lable(j,1)=PredClass_CF_DT(j);
26 |         end
27 |     end
28 |     fitness(i)=sum(particle_lable == testLabels)/size(testLabels,1 );
29 | end


--------------------------------------------------------------------------------
/ensemble_PSO.m:
--------------------------------------------------------------------------------
 1 | function [gbest,gbestFit]=ensemble_PSO(PredClass_SVM,PredClass_KNN,PredClass_NB,traindata,trainLabels)
 2 | 
 3 | N=25;
 4 | dim=3;
 5 | population=rand(N,dim);
 6 | 
 7 | [fitness]=Fitness_func(PredClass_SVM,PredClass_KNN,PredClass_NB,N,dim,traindata,trainLabels,population);
 8 | 
 9 | pbest=population;
10 | [gbestFit,indx]=max(fitness);
11 | 
12 | pbestFit = fitness;
13 | gbest = population(indx , :);
14 | 
15 | 
16 | t=0;
17 | iteration=50;
18 | vmax = 6;
19 | vmin = -6;
20 | velocity = -6+ rand(size(population))*12;
21 | w=0.7298;
22 | c1=1.5;
23 | c2=1.5;
24 | 
25 | while t<iteration
26 |     t=t+1;
27 |     
28 |     for a=1:N
29 |         r1=rand;
30 |         r2=rand;
31 |         for b=1:dim
32 |             update_velocity(a,b)=w*velocity(a,b)+c1*r1*(pbest(a,b)-population(a,b))-c2*r2*(gbest(1,b)-population(a,b));
33 |         end
34 |     end
35 |     
36 |             
37 | velocity = update_velocity;
38 | for s=1:N
39 |     for u=1:dim
40 |         update_population(s,u)=population(s,u)+velocity(s,u);
41 |         if  update_population(s,u)>=1 ||  update_population(s,u)<=0
42 |              update_population(s,u)=population(s,u);
43 |         end
44 |             
45 |     end
46 | end
47 | 
48 | 
49 | 
50 | population=update_population;
51 | 
52 | [fitness]=Fitness_func(PredClass_SVM,PredClass_KNN,PredClass_NB,N,dim,traindata,trainLabels,population);
53 | 
54 | for m=1:N
55 |     if pbestFit(1,m) < fitness(1,m)
56 |         
57 |         pbestFit(1,m)=fitness(1,m);
58 |         pbest(m,:)=population(m,:);
59 |         
60 |     end
61 | end
62 | 
63 | [update_gbestFit,update_indx]=max(fitness);
64 | 
65 | if gbestFit < update_gbestFit
66 |     gbestFit = update_gbestFit;
67 |     
68 |     indx = update_indx;
69 |     gbest = population(indx , :);
70 | end
71 | 
72 | end
73 | 


--------------------------------------------------------------------------------