蚁群聚类程序.docVIP

% Ant Colony Optimization for the Clustering clc; clear; N =44;% N = number_of_test_sample; n =3;% n = number_of_attribute_of_test_sample; % K = number_of_cluster; K = 3; % R = number_of_ants; R = 10; % t_max = MaxIterations; t_max = 1000; % X = test_sample_matrix; xdata=xlsread(2011.xls); X=zscore(xdata); % INITIALIZATION c = 10^-2;% pheromone trails tau = ones(N,K) * c; %tau = [ % 0. 0. 0.; % 0. 0. 0.; % 0. 0. 0.; % 0. 0. 0.; % 0. 0. 0.; % 0. 0. 0.; % 0. 0. 0.; % 0. 0. 0.; %]; q = 0.9;% probability threshold q rho = 0.1;% evaporation rate best_solution_function_value = inf;% fitness function t = 1; while ((t = t_max)) % SEND R ANTS EACH WITH EMPTY SOLUTION STRING,S % ============================= solution_string = zeros(R,N+1);% solution string : every ant construction solution and fitness function % COMPUTE R SOLUTION for i = 1 : R % solution_string(1,1:N) = randint(1,N,[1,K]); r = rand(1,N); %construct solution Si using pheromone trail for g = 1 : N if r(g) q tau_max = max(tau(g,:)); Cluster_number = find(tau(g,:)==tau_max); solution_string(i,g) = Cluster_number(1); %%%%%%%%%%%%% 1--i else sum_p = sum(tau(g,:)); p = tau(g,:) / sum_p; for u = 2 : K p(u) = p(u) + p(u-1); end rr = rand; %Cluster_number = K; for s = 1 : K if (rr = p(s)) Cluster_number = s; solution_string(i,g) = Cluster_number; % 1--i break; end end end end % compute weights(weight) of all test samples and cluster % centers(cluster_center) weight = zeros(N,K); for h = 1:N Cluster_index = solution_string(i,h); %%%%%%%%%%%%% 1--i weight(h,Cluster_index) = 1; end cluster_center = zeros(K,n); for j = 1:K for v = 1:n