机器学习|基于SVM的鸢尾花数据集分类实现

2023年4月11日上午2:10 • 机器学习

　　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(:,1:4);%特征列表
classlabel=data(:,5);%对应类别
n = randperm(size(features,1));%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(:,5)==0);
class_1 = find(data(:,5)==1);
class_2 = find(data(:,5)==2);%返回类别为2的位置索引
subplot(3,2,1)
hold on
scatter(features(class_0,1),features(class_0,2),'x','b')
scatter(features(class_1,1),features(class_1,2),'+','g')
scatter(features(class_2,1),features(class_2,2),'o','r')
subplot(3,2,2)
hold on
scatter(features(class_0,1),features(class_0,3),'x','b')
scatter(features(class_1,1),features(class_1,3),'+','g')
scatter(features(class_2,1),features(class_2,3),'o','r')
subplot(3,2,3)
hold on
scatter(features(class_0,1),features(class_0,4),'x','b')
scatter(features(class_1,1),features(class_1,4),'+','g')
scatter(features(class_2,1),features(class_2,4),'o','r')
subplot(3,2,4)
hold on
scatter(features(class_0,2),features(class_0,3),'x','b')
scatter(features(class_1,2),features(class_1,3),'+','g')
scatter(features(class_2,2),features(class_2,3),'o','r')
subplot(3,2,5)
hold on
scatter(features(class_0,2),features(class_0,4),'x','b')
scatter(features(class_1,2),features(class_1,4),'+','g')
scatter(features(class_2,2),features(class_2,4),'o','r')
subplot(3,2,6)
hold on
scatter(features(class_0,3),features(class_0,4),'x','b')
scatter(features(class_1,3),features(class_1,4),'+','g')
scatter(features(class_2,3),features(class_2,4),'o','r')

机器学习|基于SVM的鸢尾花数据集分类实现

　　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(1:70),:);
train_label=classlabel(n(1:70),:);
%% 测试集--30个样本
test_features=features(n(71:end),:);
test_label=classlabel(n(71: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,1)*100; 
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,1)*100;
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