# 【lssvm回归预测】基于鸽群算法优化最小二乘支持向量机PIO-lssvm实现数据回归预测附matlab代码

✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，matlab项目合作可私信。

🍎个人主页：Matlab科研工作室

🍊个人信条：格物致知。

## ⛄ 内容介绍

LSSVM 模型中的参数选择对模型的影响较大,采用鸽群优化算法进行模型参数的全局选优,用历史负荷数据和天气气象因素作为输入,建立优化电力负荷预测模型进行仿真.利用 PIO- LSSVM 模型对华东某市电力负荷进行验证分析.实验结果表明:鸽群算法优化的LSSVM 模型相比 LSSVM 具有更高的预测精度.

## ⛄ 部分代码

function [Best_vulture1_F,Best_vulture1_X,convergence_curve]=PIO(pigeonnum,max_iter,lower_bound,upper_bound,D,fobj);

%%####A new bio-inspired swarm intelligence optimizer ?C pigeon inspired

%%####optimization (PIO) is presented by simulating pigeons homing

%%####behaviors. Homing pigeons can easily find their homes by using

%%####three homing tools: magnetic field, sun and landmarks. In this

%%#### newly invented algorithm, map and compass operator model is

%%#### presented based on magnetic field and sun, while landmark operator

%%#### model is presented based on landmarks. For some tough functions,

%%#### it can quickly find the optimum, and it performs powerfully. For the

%%#### most important reason, it combines some advantages of algorithms

%%#### such as particle swarm optimization and artificial fish school algorithm.

%***************initialization*******************

T1=max_iter;     %Global search algebra

T2=max_iter;     %Local search algebra

% pigeonnum=30;    %number

% D = 30;     % dimensionality

R=0.3;     %parameters of magnetic field

bound=[lower_bound,upper_bound];    %hunting zone

tol = 1e-7;

%**************initialization of the individual pigeon************

for i=1:pigeonnum                                                           %时间复杂度O(pigeonum*D*2)

for j=1:D

x(i,j)=bound(1)+rand*(bound(2)-bound(1));

v(i,j)=rand;

end

end

%**************calculate the fitness of pigeon***********

for i=1:pigeonnum                                                           %时间复杂度O(pigeonum*2)

p(i)=fobj(x(i,:));

p_best(i,:)=x(i,:);

end

%**************find the optimal pigeons********************

g_best=x(1,:);

for i=2:pigeonnum                                                           %时间复杂度O(pigeonum-1)

if fobj(g_best)>fobj(x(i,:))

g_best=x(i,:);

end

end

%************  magnetic compass and solar operator********************

for t=1:T1                                                                  %时间复杂度O(T1*(pigeonum*(2D+5))+1)

for i=1:pigeonnum                                                       %时间复杂度O(pigeonum*(2D+5))

v(i,:)=v(i,:)+rand*(p_best(i,:)-x(i,:));

x(i,:)=x(i,:)*(1-exp(-R*t))+v(i,:);   %check whether beyond the searching space

for j=1:D                                    % magnetic field and solar operator

if abs(i-1)<=eps

if x(i,j)<bound(1)||x(i,j)>bound(2)

x(i,j)=bound(1)+rand*(bound(2)-bound(1));

v(i,j)=rand;

end

else

if x(i,j)<bound(1)||x(i,j)>bound(2)

x(i,j)=x(i-1,j);

v(i,j)=v(i-1,j);

end

end

end

if fobj(x(i,:))<p(i)                         %renewal individual fitness

p(i)=fobj(x(i,:));

p_best(i,:)=x(i,:);

end

if p(i)<fobj(g_best)                         %renewal global fitness

g_best=p_best(i,:);

end

end

result(t)=fobj(g_best);

end

%*************???????**********************

for t=1:T2                                                                  %时间复杂度O(T2*pigeonum*pigeonum)

for i=1:pigeonnum-1                             %sort the pigeons

for j=i+1:pigeonnum

if fobj(x(i,:))>fobj(x(j,:))

temp_pigeon=x(i,:);

x(i,:)=x(j,:);

x(j,:)=temp_pigeon;

end

end

end

pigeonnum=ceil(pigeonnum/2);               %remove half of the pigeons according to the landmark

for i=1:pigeonnum

p(i)=fobj(x(i,:));                     %calculate fitness and location of the pigeon after sorting

p_best(i,:)=x(i,:);

end

for i=1:pigeonnum                                %local searching

for j=1:D                                    %check whether beyond the searching space

x(i,j) = x(i,j) + rand*(pigeoncenter(j)-x(i,j));

while x(i,j)<bound(1)||x(i,j)>bound(2)

x(i,j) = x(i,j) + rand*(pigeoncenter(j)-x(i,j));

end

end

if fobj(x(i,:))<p(i)                         %renewal individual fitness

p(i)=fobj(x(i,:));

p_best(i,:)=x(i,:);

end

if p(i)<fobj(g_best)                         %renewal global fitness

g_best=p_best(i,:);

end

end

result(t+T1)=fobj(g_best);

end

% figure                                               %graph

% for t=1:T1+T2-1

%     plot([t,t+1],[result(t),result(t+1)]);

%     hold on;

% end

% xlabel('Iterative curve');

% ylabel('Function value');

% title(['The best value is ' num2str()])

[Best_vulture1_F,indx]=min(result);

Best_vulture1_X=g_best;

convergence_curve=result;

end

## ⛄ 参考文献

[1]鞠彬王嘉毅. 基于粒子群算法与最小二乘支持向量机的ET0模拟[J]. 水资源保护, 2016, 32(4):74-79.

[2]吴文江, 陈其工, and 高文根. "基于 PSO 优化参数的最小二乘支持向量机短期负荷预测." 重庆理工大学学报（自然科学版） 030.003(2016):112-115.

## ⛄ Matlab代码关注

##### ❤️ 关注我领取海量matlab电子书和数学建模资料

|
2天前
|

m基于GA-GRU遗传优化门控循环单元网络的电力负荷数据预测算法matlab仿真

13 4
|
2天前
|

13 1
|
4天前
|

22 5
|
4天前
|

20 1
|
11天前
|

20 0
|
11天前
|

18 1
|
11天前

50 1
|
11天前
|

35 0
|
11天前
|

20 1
|
11天前
|

21 0