|
EDA365欢迎您登录!
您需要 登录 才可以下载或查看,没有帐号?注册
x
8 W. @# l0 U' ?+ Y7 A t1 r
NSGA2算法特征选择MATLAB实现(多目标)
3 i4 s5 a% V! {$ ~
7 v) J, K1 T5 G% V4 ^, N0 J
! d: s- S/ c7 H8 u1 T k利用nsga2进行进行特征选择其主要思想是:将子集的选择看作是一个搜索寻优问题(wrapper方法),生成不同的组合,对组合进行评价,再与其他的组合进行比较。这样就将子集的选择看作是一个是一个优化问题。/ D. F8 j+ I4 ^; @3 p+ L
. g6 c1 W& N4 q5 l0 P需要优化的两个目标为特征数和精度。
$ q/ N7 m6 M' P7 N D7 N5 H, n1 R7 ~ O! V
nsga2是一个多目标优化算法。5 A/ W" f% U9 Y3 x9 R# s
8 c) y7 t% P7 f7 m: ^4 U具体的nsga2通用算法请看:NSGA2算法MATLAB实现(能够自定义优化函数)
4 x! G7 N2 e0 _6 V3 ~. l( x' E2 f, y, \: D1 w! Z6 Q1 B' J# r
具体的特征选择代码在上述代码的基础上改了两个①主函数②评价函数,增加了一个数据分成训练集和测试集的函数:
, M% t, y5 y) q2 R) a& D8 k' B2 w2 h' h- J
- function divide_datasets()
- load Parkinson.mat;
- dataMat=Parkinson_f;
- len=size(dataMat,1);
- %归一化
- maxV = max(dataMat);
- minV = min(dataMat);
- range = maxV-minV;
- newdataMat = (dataMat-repmat(minV,[len,1]))./(repmat(range,[len,1]));
- Indices = crossvalind('Kfold', length(Parkinson_label), 10);
- site = find(Indices==1|Indices==2|Indices==3);
- train_F = newdataMat(site,:);
- train_L = Parkinson_label(site);
- site2 = find(Indices~=1&Indices~=2&Indices~=3);
- test_F = newdataMat(site2,:);
- test_L =Parkinson_label(site2);
- save train_F train_F;
- save train_L train_L;
- save test_F test_F;
- save test_L test_L;
- end
- %what doesn't kill you makes you stronger, stand a little taller,doesn't mean i'm over cause you're gonw.- s# V& r* e; N& b# v( ]' R. z
. r( Q7 J$ X: l4 X2 K- y, p0 p9 v$ E0 C) s9 X" l3 P5 N5 z
MATLAB代码主函数:
, l* C+ z$ ~, W$ [# T9 `0 [
7 z) H: b8 o/ U, N x% o- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
- %此处可以更改
- %更多机器学习内容请访问omegaxyz.com
- clc;
- clear;
- pop = 500; %种群数量
- gen = 100; %迭代次数
- M = 2; %目标数量
- V = 22; %维度
- min_range = zeros(1, V); %下界
- max_range = ones(1,V); %上界
- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
- %特征选择
- divide_datasets();
- global answer
- answer=cell(M,3);
- global choice %选出的特征个数
- choice=0.8;
- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
- chromosome = initialize_variables(pop, M, V, min_range, max_range);
- chromosome = non_domination_sort_mod(chromosome, M, V);
- for i = 1 : gen
- pool = round(pop/2);
- tour = 2;
- parent_chromosome = tournament_selection(chromosome, pool, tour);
- mu = 20;
- mum = 20;
- offspring_chromosome = genetic_operator(parent_chromosome,M, V, mu, mum, min_range, max_range);
- [main_pop,~] = size(chromosome);
- [offspring_pop,~] = size(offspring_chromosome);
- clear temp
- intermediate_chromosome(1:main_pop,:) = chromosome;
- intermediate_chromosome(main_pop + 1 : main_pop + offspring_pop,1 : M+V) = offspring_chromosome;
- intermediate_chromosome = non_domination_sort_mod(intermediate_chromosome, M, V);
- chromosome = replace_chromosome(intermediate_chromosome, M, V, pop);
- if ~mod(i,100)
- clc;
- fprintf('%d generations completed\n',i);
- end
- end
- if M == 2
- plot(chromosome(:,V + 1),chromosome(:,V + 2),'*');
- xlabel('f_1'); ylabel('f_2');
- title('Pareto Optimal Front');
- elseif M == 3
- plot3(chromosome(:,V + 1),chromosome(:,V + 2),chromosome(:,V + 3),'*');
- xlabel('f_1'); ylabel('f_2'); zlabel('f_3');
- title('Pareto Optimal SuRFace');
- end+ p" d' V3 @" m4 ] N8 F
) ]1 o9 n( D6 O: n$ Z+ Y
评价函数(利用林志仁SVM进行训练):+ z+ Y) C# Z- n
; B, T5 f6 x q' F! e- function f = evaluate_objective(x, M, V, i)
- f = [];
- global answer
- global choice
- load train_F.mat;
- load train_L.mat;
- load test_F.mat;
- load test_L.mat;
- temp_x = x(1:V);
- inmodel = temp_x>choice;%%%%%设定恰当的阈值选择特征
- f(1) = sum(inmodel(1,:));
- answer(i,1)={f(1)};
- model = libsvmtrain(train_L,train_F(:,inmodel), '-s 0 -t 2 -c 1.2 -g 2.8');
- [predict_label, ~, ~] = libsvmpredict(test_L,test_F(:,inmodel),model,'-q');
- error=0;
- for j=1:length(test_L)
- if(predict_label(j,1) ~= test_L(j,1))
- error = error+1;
- end
- end
- error = error/length(test_L);
- f(2) = error;
- answer(i,2)={error};
- answer(i,3)={inmodel};
- end/ ~+ @" G( i. W
3 T5 v! ~6 a' q1 E1 t: E5 s6 w7 ]
选的的数据集请从UCI上下载。; n8 J1 t6 l+ E/ G8 u) a
, X+ \2 G% v) i5 h结果:
' @/ C2 l: ]& L+ P2 G
$ K3 |/ q1 `, g①pareto面3 F7 z( J! K2 M) ?3 l0 l8 H
* G% m4 g4 \& ^6 d1 y; L
+ R- q6 K, D/ l7 n
' j% X) m, X+ q$ F/ U* a7 L
最后粒子的数据(选出的特征数和精确度)/ Z$ y( o. E$ p/ K6 Q
6 P. y7 }% R H5 ^8 q1 Y: ^
* I' L% u0 \1 m& L1 ~4 b; C0 E. D1 w3 x
; d- \: x2 ~' R) P: `5 ?5 A |
|