JAWABAN UJIAN AKHIR SEMESTER
NEURAL NETWORK
MUHAMMAD YUDHA SYUHADA
167038075
PROGRAM STUDI MAGISTER TEKNIK INFORMATIKA
FAKULTAS ILMU KOMPUTER DAN TEKNOLOGI INFORMASI
UNIVERSITAS SUMATERA UTARA
MEDAN
2017
1. Source code matlab dengan inputan 5 layer (X1, X2, X3, X4, X5), 1 hidden
layer dan 1 output layer
% Mempersiapkan data latih dan target latih
% data latih berupa matriks (X1;X2;X3;X4;X5) => 5 x 50
data_latih =
[90,85,80,85,85,85,90,85,85,80,70,90,85,80,85,85,85,90,85,85,80,70,90,85,80
,85,85,85,80,80,85,90,85,85,80,70,90,85,80,85,85,85,80,80,85,70,80,85,91,85
;...
85,85,85,90,90,90,85,95,85,90,80,85,85,85,90,90,90,85,95,85,90,80,85,85,85,
90,90,90,85,85,90,85,95,85,90,80,85,85,85,90,90,90,85,85,85,20,85,90,85,95;
...
90,85,80,85,90,80,85,85,85,90,75,90,78,89,76,65,56,87,45,45,98,75,90,90,43,
34,56,49,85,75,78,56,89,93,76,75,96,78,66,76,89,90,85,67,78,87,90,78,74,58;
...
70,78,71,72,76,78,76,81,77,70,71,70,78,71,72,76,78,76,81,77,70,71,70,78,71,
72,76,74,71,70,78,76,81,77,70,71,70,78,71,72,76,74,71,70,73,70,80,78,76,81;
...
66,61,66,60,57,66,67,62,65,70,78,66,61,66,60,57,66,67,62,65,70,78,66,61,66,
60,57,54,60,58,66,67,62,65,70,78,66,61,66,60,57,54,60,58,54,30,70,66,67,62]
;
% isi target dengan menjadi matriks 1 x 50
target_latih =
[74,73,72,72,72,75,76,76,75,75,75,74,73,73,71,70,73,76,72,71,76,75,74,74,69
,67,69,66,70,68,75,73,76,75,74,75,75,73,71,71,72,70,70,67,68,51,78,75,75,73
];
[~,N] = size(data_latih);
% Pembuatan JST
%[1 1] menunjukkan hidden layer 1 dan output 1
net = newff(minmax(data_latih),[1 1],{'logsig','purelin'},'traingdx');
%inisialisasi bobot hidden layer
net.IW{1,1} = [-7.62,0.97,-2.60,-9.55,0.5];
%inisialisasi bobot output layer
net.LW{2,1} = [-2.40];
%inisialisasi bias
net.b{1,1} = [1];
%inisialisasi bias yang langsung ke output layer
net.b{2,1} = [1];
% Memberikan nilai untuk mempengaruhi proses pelatihan
net.performFcn = 'mse';
net.trainParam.goal = 0.01;
net.trainParam.show = 20;
net.trainParam.epochs = 1000;
net.trainParam.mc = 0.95;
net.trainParam.lr = 0.1;
% Proses training
[net_keluaran,tr,Y,E] = train(net,data_latih,target_latih);
% Hasil setelah pelatihan
bobot_hidden = net_keluaran.IW{1,1};
bobot_keluaran = net_keluaran.LW{2,1};
bias_hidden = net_keluaran.b{1,1};
bias_keluaran = net_keluaran.b{2,1};
jumlah_iterasi = tr.num_epochs;
nilai_keluaran = Y;
nilai_error = E;
error_MSE = (1/N)*sum(nilai_error.^2);
bobot_hidden
bobot_keluaran
bias_hidden
bias_keluaran
2. Hasil bobot dari pelatihan
 W1 = -7.6200
 W2 = 0.9700
 W3 = -2.6000
 W4 = -9.5500
 W5 = 0.5000
 W6 = 1
 W7 = 72.2600
 W8 = -2.4000
3. Gambar / Screenshot