cannot get good result by newff

조회 수: 8 (최근 30일)
reza
reza 2012년 2월 23일
Hi everyone, i have 12 years daily data,and i'm trying to use newff to forecast daily, monthly.i define my data as first 3 years as input and forth years as target.it means i have a (1905:4) matrix.and i'm using nn code as below, but unfortunately i don't get good result.as i'm new in matlab cannot figure out the problem.please someone help me(new code,new advice,...). and another question is how should i define my data for monthly forecasting. thanks reza
my code is:
Close all;
clear all;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% to create the data for the model"
% and preprocess it to be compatable with neural network%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
load Input
num=data;
input1=num(:,1);
input2=num(:,2);
input3=num(:,3);
t=num(:,4);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% this part to start the trainign session for the ANN model%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
p1=num(1:750,(1:3)); p11=p1';
A =t(1:750,1); A1=A';
net_gas = newff([ 0 1;0 1;0 1],[10,25,1],{'tansig','tansig','purelin'},'trainlm');
net_gas.trainParam.show = 100;
net_gas.trainParam.epochs = 150;
%net5.performFcn = 'msereg';
%net5.performParam.ratio = 0.1;
net_gas.trainParam.goal = 1e-10;
%training session
net_gas=train(net_gas,p11,A1);
save net_gas
%Simulation session
y1 = sim(net_gas,p11);
y1f= y1;
H1=y1f';
%H2=H1*388.8;
%Neural Network Performance
Error= (H1-dm(1:750))./dm(1:750)*100;
figure
x= 1:1:750;
plot (x, Error); grid
xlabel('Expr#');
ylabel('%Error');
title('NN performance for H');
axis([1 750 -25 25]);
%Test session
pt1=num((751:1095),(1:3));
pt11=pt1';
yt1 = sim(net_gas,pt11);
yt1f= yt1;
Ht1=yt1f';
%Ht2= Ht1*388.8;
Error_test= (Ht1-dm(751:1095))./dm(751:1095)*100;
figure
xt= 1:1:365;
plot (xt, Error_test); grid
xlabel('Exp#');
ylabel('%Error_test');
title('NN performance for H');
axis([1 365 -100 100]);

채택된 답변

Greg Heath
Greg Heath 2012년 2월 24일
See my answer in the newsgroup.
Greg

추가 답변 (0개)

카테고리

Help CenterFile Exchange에서 Deep Learning Toolbox에 대해 자세히 알아보기

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by