Matlab实现WOA-BiLSTM鲸鱼算法优化双向长短期记忆网络时间序列预测（完整程序和数据）

%% 主程序 WOA-BiLSTM，鲸鱼算法优化双向长短期记忆神经网络预测 % 初始化程序 clc;clear;close all;format compact %% data = xlsread('data.xlsx', 'Sheet1'); n=7; [x,y]=data_process(data,n);%前n个时刻 预测下一个时刻 input =x; % 输入数据 output=y; % 输出数据 % load analysis modes % 预设鲸鱼优化算法的参数 popsize = 5; %种群数量 maxgen = 20; % WOA最大的迭代次数 dim = 3; % 优化的自变量个数 ，依次为BiLSTM的第一和第二隐含层节点数，最大训练代数，初始学习率 %输入输出数据 L = size(output, 1); %样本总数 % testNumber = round(L*0.2); %设置四个样本测试 testNumber =20; %设置四个样本测试 trainNumber = L - testNumber; %训练样本数量 % 划分训练集和测试集 p_train = input(1:trainNumber, :)'; t_train = output(1:trainNumber, :)'; % 测试集 p_test = input(trainNumber+1:trainNumber+testNumber,:)'; t_test = output(trainNumber+1:trainNumber+testNumber,:)'; %% 归一化和元胞数组处理 [pn_train ,ps]= mapminmax(p_train); [tn_train, ts] = mapminmax(t_train); %测试集的归一化 pn_test = mapminmax('apply', p_test, ps); tn_test = mapminmax('apply', t_test, ts); for i = 1:length(t_train) P_train{i,1} = pn_train(:,i); end for i = 1:length(t_test) P_test{i,1} = pn_test(:,i); end %% 归一化和元胞数组处理p 4; % 优化的自变量个数 ，依次为BiLSTM的第一和第二隐含层节点数，最大训练代数，初始学习率 lb = [1, 1, 0.001]; % 下限 ub = [10,20, 0.01]; % 上限 % initialize position vector and score for the leader，初始化位置向量和领导者得分 Leader_pos=zeros(1,dim); Leader_score=inf; %change this to -inf for maximization problems，将此更改为-inf以获得最大化问题，Inf无穷大 %Initialize the positions of search agents Positions=initialization(popsize,dim,ub,lb);%Positions，存放数个个体的多维位置。 Convergence_curve=zeros(1,maxgen);%Convergence_curve收敛曲线 t=1;% Loop counter % Main loop while t<maxgen+1 % 首先 迭代次数的while循环 为2 也就是鲸鱼群体会进化2代 disp(['current iteration is: ', num2str(t)]) for i=1:size(Positions,1)%对每个个体一个一个检查是否越界 % 种群规模是2 所以嵌套了2次训练 % Return back the search agents that go beyond the boundaries of % the search space，返回超出搜索空间边界的搜索代理 Flag4ub=Positions(i,:)>ub; Flag4lb=Positions(i,:)<lb; Positions(i,:)=(Positions(i,:).*(~(Flag4ub+Flag4lb)))+ub.*Flag4ub+lb.*Flag4lb;%超过最大值的设置成最大值，超过最小值的设置成最小值 % 评估个体适应度，会调用func.m子函数，把个体位置坐标赋给BiLSTM参数中，执行的训练 % Calculate objective function for each search agent，目标函数值的计算 [fitness,net]=func(Positions(i,:),p_train,P_train,P_test, tn_train,ts,t_train,t_test); if fitness<Leader_score % Change this to > for maximization problem Leader_score=fitness; % Update alpha Leader_pos=Positions(i,:); net1 = net; end end a=2-t*((2)/maxgen); % a decreases linearly fron 2 to 0 in Eq. (2.3) % a2 linearly dicreases from -1 to -2 to calculate t in Eq. (3.12)，有疑问？ a2=-1+t*((-1)/maxgen); % Update the Position of search agents，参数更新 for i=1:size(Positions,1) r1=rand(); % r1 is a random number in [0,1] r2=rand(); % r2 is a random number in [0,1] A=2*a*r1-a; % Eq. (2.3) in the paper C=2*r2; % Eq. (2.4) in the paper b=1; % parameters in Eq. (2.5) l=(a2-1)*rand+1; % parameters in Eq. (2.5) p = rand(); % p in Eq. (2.6) for j=1:size(Positions,2)%对每一个个体地多维度进行循环运算 if p<0.5%收缩包围机制 if abs(A)>=1 rand_leader_index = floor(popsize*rand()+1);%floor将 X 的每个元素四舍五入到小于或等于该元素的最接近整数 X_rand = Positions(rand_leader_index, :); D_X_rand=abs(C*X_rand(j)-Positions(i,j)); % Eq. (2.7) Positions(i,j)=X_rand(j)-A*D_X_rand; % Eq. (2.8) elseif abs(A)<1 D_Leader=abs(C*Leader_pos(j)-Positions(i,j)); % Eq. (2.1) Positions(i,j)=Leader_pos(j)-A*D_Leader; % Eq. (2.2) end elseif p>=0.5%螺旋更新位置 distance2Leader=abs(Leader_pos(j)-Positions(i,j)); % Eq. (2.5) Positions(i,j)=distance2Leader*exp(b.*l).*cos(l.*2*pi)+Leader_pos(j); end end end Convergence_curve(t)=Leader_score; t=t+1; disp(t) end figure; plot(Convergence_curve,'b-','LineWidth',1.0) grid on xlabel('进化代数') ylabel('最佳适应度') title('WOA优化过程曲线') disp('WOA优化BiLSTM的结果:') disp(strcat('最优参数: ', num2str([round(Leader_pos(1:2)),Leader_pos(3)]))) disp(strcat('最佳适应度: ', num2str(Leader_score))) %% 将WOA优化后的参数赋给BiLSTM神经网络 while循环结束之后 优化完的最佳参数付给BiLSTM 再次训练 % 参数设置 numFeatures = size(p_train, 1); %输入层节点 numHiddenUnits1 = round(Leader_pos(1)); % 第一隐含层节点 numResponses = 1; %全连接层节点（等于标签的数量） %% 创建网络 layers = [ ... sequenceInputLayer(numFeatures) bilstmLayer(numHiddenUnits1,'OutputMode','last','name','hidden1') dropoutLayer(0.2,'name','dropout_1') %隐藏层1权重丢失率，防止过拟合 fullyConnectedLayer(numResponses,'name','fullconnect') regressionLayer('name','out')]; % %回归层 %% 参数设定 %指定训练选项，求解器设置为adam options = trainingOptions('adam', ... %优化算法 'MaxEpochs',round(Leader_pos(2)), ... %遍历样本最大循环数 'GradientThreshold',1, ... %梯度阈值 'InitialLearnRate',Leader_pos(3), ... %初始学习率 'LearnRateSchedule','piecewise', ... %学习率计划 'LearnRateDropPeriod',100, ... %50个epoch后学习率更新 'LearnRateDropFactor',0.01, ... % 通过乘以因子 0.1 来降低学习率 'MiniBatchSize',100,... % 批处理样本大小 'Verbose',1, ... %命令控制台是否打印训练过程 'Plots','training-progress'); %% 训练BiLSTM % net = trainNetwork(P_train,tn_train',layers,options); net = net1; %% 测试集预测 testn_simu = predict(net,P_test); test_simu= mapminmax('reverse',testn_simu', ts); disp(' ') disp('WOA-BiLSTM神经网络预测性能分析:') [e, ape]=caculate_perf(t_test, test_simu); rmse=sqrt(mean( (test_simu - t_test).^2)); %% 作图分析 %%真实值与预测值误差比较 %----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- figure plot(test_simu,'-s','Color',[0 0 255]./255,'linewidth',1,'Markersize',5,'MarkerFaceColor',[0 0 255]./255) hold on plot(t_test,'-o','Color',[0 0 0]./255,'linewidth',0.8,'Markersize',4,'MarkerFaceColor',[0 0 0]./255) legend('WOA-BiLSTM预测测试数据','实际分析数据','Location','NorthWest','FontName','华文宋体'); title('WOA-BiLSTM模型预测结果及真实值','fontsize',12,'FontName','华文宋体') xlabel('样本','fontsize',12,'FontName','华文宋体'); ylabel('数值','fontsize',12,'FontName','华文宋体'); %----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- figure bar((test_simu - t_test)) legend('WOA-BiLSTM模型测试集误差','Locatio