【预测模型-ANN预测】基于布谷鸟优化人工神经网络算法实现数据预测附matlab代码.zip

% ----------------------------------------------------------------- % Cuckoo Optimization Algorithm (COA) by Ramin Rajabioun % % Programmed by Ramin Rajabioun % % ----------------------------------------------------------------- % Paper: R. Rajabioun. Cuckoo Optimization Algorithm, Applied Soft % Computing 11 (2011) 5508?5518 % ----------------------------------------------------------------% % This program implements a standard version of Cuckoo % % Optimization Algorithm (COA) which minimizes any Cost Function % % % To use the code easily prepare your cost function and type its % % name in: costFunction = 'YourCostFunctionName', then set number % % of optimization parameters in "npar" and set the upper and lower% % bands of the problem % % --------------------------------------------------------------- % clc, clear, close all global input output numOfNeurons NumofInputs %% Set problem parameters % select a cost function: costFunction = 'ANN_Cost'; input = (-6:0.5:6)'; output = sin(input+eps)./(input+eps); %% ANN Structure numOfNeurons = 3; NumofInputs = 1; npar = (NumofInputs+1) * numOfNeurons + (numOfNeurons+1); % Number of Optimization Parameters varLo = -20; % Lower band of parameter varHi = 20; % Higher band of parameter %% Set COA parameters numCuckooS = 5; % number of initial population minNumberOfEggs = 2; % minimum number of eggs for each cuckoo maxNumberOfEggs = 4; % maximum number of eggs for each cuckoo maxIter = 200; % maximum iterations of the Cuckoo Algorithm knnClusterNum = 1; % number of clusters that we want to make motionCoeff = 20; % Lambda variable in COA paper, default=2 accuracy = 0e-10; % How much accuracy in answer is needed maxNumOfCuckoos = 20; % maximum number of cuckoos that can live at the same time radiusCoeff = 0.05; % Control parameter of egg laying cuckooPopVariance = 1e-10; % population variance that cuts the optimization %% initialize population: cuckooPop = cell(numCuckooS,1); % initialize egg laying center for each cuckoo for cuckooNumber = 1:numCuckooS cuckooPop{cuckooNumber}.center = ( varHi-varLo )*rand(1,npar) + varLo; end %% Start Cuckoo Optimization Algorithm iteration = 0; maxProfit = -1e20; % Let initial value be negative number goalPoint = (varHi - varLo)*rand(1,npar) + varLo; % a random goalpoint to start COA globalBestCuckoo = goalPoint; globalMaxProfit = maxProfit; profitVector = []; while ( (iteration <= maxIter) && (-maxProfit > accuracy) ) iteration = iteration + 1 % initialize number of eggs for each cuckoo for cuckooNumber = 1:numCuckooS cuckooPop{cuckooNumber}.numberOfEggs = floor( (maxNumberOfEggs - minNumberOfEggs) * rand + minNumberOfEggs ); end % get total number of available eggs between all cuckooS summ = 0; for cuckooNumber = 1:numCuckooS summ = summ + cuckooPop{cuckooNumber}.numberOfEggs; end % calculate egg laying radius for each Cuckoo, considering problem % limitations and ratio of each cuckoo's eggs for cuckooNumber = 1:numCuckooS cuckooPop{cuckooNumber}.eggLayingRadius = cuckooPop{cuckooNumber}.numberOfEggs/summ * ( radiusCoeff * (varHi-varLo) ); end % To lay eggs, we produced some radius values less than egg laying % radius of each cuckoo for cuckooNumber = 1:numCuckooS cuckooPop{cuckooNumber}.eggLayingRadiuses = cuckooPop{cuckooNumber}.eggLayingRadius * rand(cuckooPop{cuckooNumber}.numberOfEggs,1); end for cuckooNumber = 1:numCuckooS params = cuckooPop{cuckooNumber}.center; % get center values tmpRadiuses = cuckooPop{cuckooNumber}.eggLayingRadiuses; numRadiuses = numel(tmpRadiuses); % divide a (hyper)circle to 'numRadiuses' segments % This is to search all over the current habitat angles = linspace(0,2*pi,numRadiuses); % in Radians newParams = []; for cnt = 1:numRadiuses addingValue = zeros(1,npar); for iii = 1:npar randNum = floor(2*rand)+1; addingValue(iii) = (-1)^randNum * tmpRadiuses(cnt)*cos(angles(cnt)) + tmpRadiuses(cnt)*sin(angles(cnt)); end newParams = [newParams; params + addingValue ]; end % check for variable limits newParams(find(newParams>varHi)) = varHi; newParams(find(newParams<varLo)) = varLo; cuckooPop{cuckooNumber}.newPosition4Egg = newParams; end % Now egg laying is done % Now that egg positions are found, they are laid, and so its time to % remove the eggs on the same positions (because each egg only can go to one nest) for cuckooNumber = 1:numCuckooS tmpPopulation = cuckooPop{cuckooNumber}.newPosition4Egg; tmpPopulation = floor(tmpPopulation * 1e4)/1e4; ii = 2; cntt = 1; while ii <= size(tmpPopulation,1) || cntt <= size(tmpPopulation,1) if all((tmpPopulation(cntt,:) == tmpPopulation(ii,:))) tmpPopulation(ii,:) = []; end ii = ii + 1; if ii > size(tmpPopulation,1) && cntt <= size(tmpPopulation,1) cntt = cntt + 1; ii = cntt + 1; if ii > size(tmpPopulation,1) break end end end cuckooPop{cuckooNumber}.newPosition4Egg = tmpPopulation; end % Now we evalute egg positions for cuckooNumber = 1:numCuckooS cuckooPop{cuckooNumber}.profitValues = -feval(costFunction,[cuckooPop{cuckooNumber}.center ; cuckooPop{cuckooNumber}.newPosition4Egg]); end % Now we check to see if cuckoo population is more than maxNumOfCuckoos % this case we should keep 1st maxNumOfCuckoos cuckoos and kill the others allPositions = []; whichCuckooPopTheEggBelongs = []; tmpProfits = []; if numCuckooS > maxNumOfCuckoos for cuckooNumber = 1:numCuckooS tmpProfits = [tmpProfits; cuckooPop{cuckooNumber}.profitValues]; allPositions = [allPositions; [cuckooPop{cuckooNumber}.center; cuckooPop{cuckooNumber}.newPosition4Egg(:,1:npar)]]; whichCuckooPopTheEggBelongs = [whichCuckooPopTheEggBelongs; cuckooNumber*ones(size(cuckooPop{cuckooNumber}.newPosition4Egg(:,1:npar),1),1) ]; end % now we sort cuckoo profits [sortedProfits, sortingIndex] = sort(tmpProfits,'descend'); % Get best cuckoo to be copied to next generation bestCuckooCenter = allPositions(sortingIndex(1),1:npar); sortedProfits = sortedProfits(1:maxNumOfCuckoos); allPositions = allPositions(sortingIndex(1:maxNumOfCuckoos),:); clear cuckooPop for ii = 1:maxNumOfCuckoos cuckooPop{ii}.newPosition4Egg = allPositions(ii,:); cuckooPop{ii}.center = allPositions(ii,:); cuckooPop{ii}.profitValues = sortedProfits(ii); end numCuckooS = maxNumOfCuckoos; else for cuckooNumber = 1:numCuckooS tmpProfits = [tmpProfits; cuckooPop{cuckooNumber}.profitValues]; allPositions = [allPositions; [cuckooPop{cuckooNumber}.center; cuckooPop{cuckooNumber}.newPosition4Egg(:,1:npar)] ]; whichCuckooPopTheEggBelongs = [whichCuckooPopTheEggBelongs; cuckooNumber*ones(size(cuckooPop{cuckooNumber}.newPosition4Egg(:,1:npar),1),1) ]; end [sortedProfits, sortingIndex] = sort(tmpProfits,'descend'); % Get best cuckoo to be copied to next generation bestCuckooCenter = allPosi