神经网络理论与Matlab7实现

p=[58478 135185 5.46 0.23 16.5 0.21 1005.3 585.44; 67884 152369 5.46 0.27 18.7 0.26 1105.6 575.03; 74462 182563 6.01 0.25 21.6 0.28 1204.6 601.23; 78345 201587 6.12 0.26 25.8 0.29 1316.5 627.89; 82067 225689 6.21 0.26 30.5 0.31 1423.5 676.95; 89403 240568 6.37 0.28 34.9 0.33 1536.2 716.32; 95933 263856 6.38 0.28 39.8 0.36 1632.6 765.24; 104790 285697 6.65 0.30 42.5 0.39 1753.2 812.22; 116694 308765 6.65 0.30 46.7 0.41 1865.5 875.26]'; t=[102569 52365 46251; 124587 60821 56245; 148792 69253 67362; 162568 79856 78165; 186592 91658 90548; 205862 99635 98758; 226598 109862 102564; 245636 120566 111257; 263595 130378 120356]'; %设定需要归一化的行序号 a=[1 2 3 5 7 8]; P=p; %对输入向量进行归一化处理 for i=1:6 P(a(i),:)=(p(a(i),:)-min(p(a(i),:)))/(max(p(a(i),:))-min(p(a(i),:))); end %对目标向量进行归一化处理 for i=1:3 T(i,:)=(t(i,:)-min(t(i,:)))/(max(t(i,:))-min(t(i,:))); end %训练样本，P_train为输入向量，T_train为目标向量 P_train=[P(:,1) P(:,2) P(:,3) P(:,4) P(:,5) P(:,6) P(:,7)]; T_train=[T(:,1) T(:,2) T(:,3) T(:,4) T(:,5) T(:,6) T(:,7)]; %测试样本，P_test为输入向量，T_test为目标向量 P_test=[P(:,8) P(:,9)]; T_test=[T(:,8) T(:,9)]; for i=1:5 net=newgrnn(P_train,T_train,i/10); %网络对训练数据的逼近 temp=sim(net,P_train); j=3*i; y_out(j-2,:)=temp(1,:); y_out(j-1,:)=temp(2,:); y_out(j,:)=temp(3,:); %网络的预测输出 temp=sim(net,P_test); y(j-2,:)=temp(1,:); y(j-1,:)=temp(2,:); y(j,:)=temp(3,:); end y1=[y_out(1,:);y_out(2,:);y_out(3,:)]; y2=[y_out(4,:);y_out(5,:);y_out(6,:)]; y3=[y_out(7,:);y_out(8,:);y_out(9,:)]; y4=[y_out(10,:);y_out(11,:);y_out(12,:)]; y5=[y_out(13,:);y_out(14,:);y_out(15,:)]; y6=[y(1,:);y(2,:);y(3,:)]; y7=[y(4,:);y(5,:);y(6,:)]; y8=[y(7,:);y(8,:);y(9,:)]; y9=[y(10,:);y(11,:);y(12,:)]; y10=[y(13,:);y(14,:);y(15,:)]; %计算逼近误差 for i=1:7 error1(i)=norm(y1(:,i)-T_train(:,i)); error2(i)=norm(y2(:,i)-T_train(:,i)); error3(i)=norm(y3(:,i)-T_train(:,i)); error4(i)=norm(y4(:,i)-T_train(:,i)); error5(i)=norm(y5(:,i)-T_train(:,i)); end %计算预测误差 for i=1:2 error6(i)=norm(y6(:,i)-T_test(:,i)); error7(i)=norm(y7(:,i)-T_test(:,i)); error8(i)=norm(y8(:,i)-T_test(:,i)); error9(i)=norm(y9(:,i)-T_test(:,i)); error10(i)=norm(y10(:,i)-T_test(:,i)); end %绘制逼近误差曲线 plot(1:7,error1,'-*'); hold on; plot(1:7,error2,'-+'); hold on; plot(1:7,error3,'-h'); hold on; plot(1:7,error4,'-d'); hold on; plot(1:7,error5,'-o'); hold off; figure; %绘制预测误差曲线 plot(1:2,error6,'-*'); hold on; plot(1:2,error7,'-+'); hold on; plot(1:2,error8,'-h'); hold on; plot(1:2,error9,'-d'); hold on; plot(1:2,error10,'-o'); hold off;