预测代码的优化问题
以下是用支持向量机预测混沌时间序列的一个代码，可是预测时间要好长，上午10:40运行，到下午16:55还是没有出结果，大家看下，可否帮忙优化下？代码如下：clc;clear all;
load data20.txt;
=size(data20);
x=zeros(nrow,3);
tao=24*6;
for i=1:nrow
x(i,1)=data20(i);
if mod(i,tao)==0
x(i,2)=fix(i/tao);
x(i,3)=24*6;
else
x(i,3)=mod(i,tao);
x(i,2)=fix(i/tao)+1;
end
end
tau=12;
m_demension=14;
% cur_day=30;
% datanum_day=144;
%preday_index=2;
p=3;
cur_index=2736;%x(t)中的t
n_used_datset=cur_index +p;
n_record=cur_index - (m_demension -1)*tau;
for i=1:n_record
for j=1:m_demension
row_t_index=cur_index -(i -1);
jindex=row_t_index-(m_demension -j)*tau;
x_jihe(2580-(i-1),j)=x(jindex,1);
end
kk=row_t_index + p;
y_jihe(2580-(i-1),1)=x(kk,1);
end
x_train=x_jihe(1:(n_record-1),:);
y_train=y_jihe(1:(n_record-1),:);
x_fina= x_jihe(n_record,:);
y_fina= y_jihe(n_record,:);
%%%%yuce
y_pre_test=0;
e=0;
gam=10;
sig2=2;
=bay_initlssvm({x_train,y_train,'f',gam,sig2,'RBF_kernel'});
=bay_optimize({x_train,y_train,'f',gam0,sig20,'RBF_kernel','preprocess'},2);
=bay_optimize({x_train,y_train,'f',gamopt,sig20,'RBF_kernel','preprocess'},3);
gamopt
sig2opt
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
x_test=x_fina;
y_test=y_fina;
%%%%%%%%%%%%%%%%%%%%
type = 'function approximation';%这个引号里面的内容说明了用于回归运算，还有另一种查手册把。
= trainlssvm({x_train,y_train,type,gamopt,sig2opt,'RBF_kernel'});
x_single_test=x_fina;
y_pre_test= simlssvm({x_train,y_train,type,gamopt,sig2opt,'RBF_kernel','preprocess'},{alpha,b},x_single_test);
actal=y_pre_test;
%fe(i)=e(i)/y_test_fgyh*100;
% FCaculate=max(e);
% Femain=sum(abs(fe))/31
%%%%%1—————输出求得的极值
% plot(1:max_iter,actal,'-')
% hold on
% plot(1:max_iter,predict_load,'-.d')
PS：x_jihe为2580x14的矩阵；y_jihe为2580x1的矩阵。谢谢大家了！急求
页:
[1]