hello matlab community ..... I'm working with a neural network capable of classifying patterns, I have no one to turn to ... my network is not ranking very well .... it classifies all results in just a standard ... and in fact I have to classify three types ... if someone can help me to improve this algortimo'll be very grateful not know where I am going wrong ... I am so frustrated right now ... thank you
clear all close all clc
n = input('enter the first layer of neurons number\n') ;
m = input('enter the number of the neurons of the second layer\n') ;
a = input('digite 1-(40),2-(45),3-(50),4-(60),5-(70),6-(80) \n') ;
p = input('enter the number of samples you want to test\n') ;
switch a
case 1
[x] = xlsread('d_40S.xlsx') ;
case 2
[x] = xlsread('d_45S.xlsx') ;
case 3
[x] = xlsread('d_50S.xlsx') ;
case 4
[x] = xlsread('d_60S.xlsx') ;
case 5
[x] = xlsread('d_70S.xlsx') ;
case 6
[x] = xlsread('d_80S.xlsx') ;
end
[y] = xlsread('target_3eS.xlsx') ;
[w,k] = size(y) ;
%grouping the data matrix and target
matriz_dados = [x y] ;
%scrambling the matrix so as to become more random
dados = permutation(matriz_dados) ;
[l,c] = size(dados) ;
%separating training samples and testing randomly
idx = randperm(l);
id_teste = (idx <= p) ;
id_treino = (idx > p) ;
%matrix with complete set of data
dado_treino = dados(id_treino,:) ; %sample for training
dado_teste = dados(id_teste,:) ; %Test sample
%samples for training
dadotreino = dado_treino(:,1:c-k)' ;
tagtreino = dado_treino(:,c-(k-1):c)' ;
%samples for validation
dadoteste = dado_teste(:,1:c-k)' ;
tagteste = dado_teste(:,c-(k-1):c)' ;
%---------------------------------------------------------
rede = network ; %network name
rede.numinputs = 1 ;%amount of input that the network layers have
rede.numlayers = 3; %number of network layers that have
rede.biasConnect = [1; 1;1] ;%Is defined now the layers that have associated bias
%1 = will bias,0 =You will not have bias
rede.inputConnect = [1 ;0 ;0] ;%connecting the inputs network layers
rede.layerConnect = [0 0 0 ; 1 0 0;0 1 0] ;%Connection between layers
rede.outputConnect = [0 0 1] ;%layers that have connection with the output
rede.inputs{1}.range = [0 1] ;% maximum and minimum value that can take the variables
t = c -(c-k) ;
rede.inputs{1}.size = c-k ;
rede.layers{1}.size = n ;
rede.layers{2}.size = m ;
rede.layers{3}.size = t ;
for j=1:t
rede.layers{j}.transferFcn ='logsig' ;
rede.layers{j}.initFcn = 'initnw' ;
end
rede.performFcn = 'mse';
rede.trainFcn = 'trainrp';
rede.trainParam.epochs = 10000 ;
rede.view
rede = init(rede) ;
train_rede = train(rede,dadotreino,tagtreino) ;
output = rede(dadoteste) ;
output_final = vec2ind(output)' ;
% Y = sim(rede, dadoteste) ; % z = vec2ind(Y)' ;
tagtest2 = vec2ind(tagteste)' ;
M = confusionmat(tagtest2,output_final)
