基于SVM的鸢尾花数据集分类实现[使用Matlab]

　　iris数据集的中文名是安德森鸢尾花卉数据集，英文全称是Anderson’s Iris data set。iris包含150个样本，对应数据集的每行数据。每行数据包含每个样本的四个特征和样本的类别信息，所以iris数据集是一个150行5列的二维表。通俗地说，iris数据集是用来给花做分类的数据集，每个样本包含了花萼长度、花萼宽度、花瓣长度、花瓣宽度四个特征（前4列），我们需要建立一个分类器，分类器可以通过样本的四个特征来判断样本属于山鸢尾、变色鸢尾还是维吉尼亚鸢尾（这三个名词都是花的品种）。

数据的获取：

file=importdata('iris.csv');%读取csv文件中从第R-1行，第C-1列的数据开始的数据
data=file.data;
features=data(:,:);%特征列表
classlabel=data(:,);%对应类别
n = randperm(size(features,));%随机产生训练集和测试集

绘制散点图查看数据：

%% 绘制散点图
class_0 = find(data(:,)==);
class_1 = find(data(:,)==);
class_2 = find(data(:,)==);%返回类别为2的位置索引
subplot(,,)
hold on
scatter(features(class_0,),features(class_0,),'x','b')
scatter(features(class_1,),features(class_1,),'+','g')
scatter(features(class_2,),features(class_2,),'o','r')
subplot(,,)
hold on
scatter(features(class_0,),features(class_0,),'x','b')
scatter(features(class_1,),features(class_1,),'+','g')
scatter(features(class_2,),features(class_2,),'o','r')
subplot(,,)
hold on
scatter(features(class_0,),features(class_0,),'x','b')
scatter(features(class_1,),features(class_1,),'+','g')
scatter(features(class_2,),features(class_2,),'o','r')
subplot(,,)
hold on
scatter(features(class_0,),features(class_0,),'x','b')
scatter(features(class_1,),features(class_1,),'+','g')
scatter(features(class_2,),features(class_2,),'o','r')
subplot(,,)
hold on
scatter(features(class_0,),features(class_0,),'x','b')
scatter(features(class_1,),features(class_1,),'+','g')
scatter(features(class_2,),features(class_2,),'o','r')
subplot(,,)
hold on
scatter(features(class_0,),features(class_0,),'x','b')
scatter(features(class_1,),features(class_1,),'+','g')
scatter(features(class_2,),features(class_2,),'o','r')

　　曲线为根据花萼长度、花萼宽度、花瓣长度、花瓣宽度之间的关系绘制的散点图。

训练集与测试集：

%% 训练集--70个样本
train_features=features(n(:),:);
train_label=classlabel(n(:),:);
%% 测试集--30个样本
test_features=features(n(:end),:);
test_label=classlabel(n(:end),:);

数据归一化：

%% 数据归一化
 [Train_features,PS] = mapminmax(train_features');
 Train_features = Train_features';
 Test_features = mapminmax('apply',test_features',PS);
 Test_features = Test_features';

使用SVM进行分类：

%% 创建/训练SVM模型
model = svmtrain(train_label,Train_features);
%% SVM仿真测试
[predict_train_label] = svmpredict(train_label,Train_features,model);
[predict_test_label] = svmpredict(test_label,Test_features,model);
%% 打印准确率
compare_train = (train_label == predict_train_label);
accuracy_train = sum(compare_train)/size(train_label,)*;
fprintf('训练集准确率：%f\n',accuracy_train)
compare_test = (test_label == predict_test_label);
accuracy_test = sum(compare_test)/size(test_label,)*;
fprintf('测试集准确率：%f\n',accuracy_test)

结果：

*
optimization finished, #iter = 18
nu = 0.668633
obj = -21.678546, rho = 0.380620
nSV = 30, nBSV = 28
*
optimization finished, #iter = 29
nu = 0.145900
obj = -3.676315, rho = -0.010665
nSV = 9, nBSV = 4
*
optimization finished, #iter = 21
nu = 0.088102
obj = -2.256080, rho = -0.133432
nSV = 7, nBSV = 2
Total nSV = 40
Accuracy = 97.1429% (68/70) (classification)
Accuracy = 97.5% (78/80) (classification)
训练集准确率：97.142857
测试集准确率：97.500000