# 导入必要的库
from sklearn import datasets
from sklearn.model_selection import train_test_split, cross_val_score, StratifiedKFold
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, classification_report
import numpy as np
# 1. 加载 Iris 数据集
iris = datasets.load_iris()
X = iris.data # 特征数据
y = iris.target # 目标标签
# 2. 使用留出法将数据集划分为训练集和测试集，留出 1/3 样本作为测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42, stratify=y)
# 3. 使用训练集训练支持向量机（SVM）分类器（SMO）
svm_model = SVC(kernel='linear', random_state=42) # 使用线性核的支持向量机
# 训练模型
svm_model.fit(X_train, y_train)
# 4. 使用五折交叉验证评估模型的性能（准确度、精度、召回率和 F1 值）
cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)
# 准确度
accuracy_scores = cross_val_score(svm_model, X, y, cv=cv, scoring='accuracy')
# 精度、召回率、F1 值
precision_scores = cross_val_score(svm_model, X, y, cv=cv, scoring='precision_weighted')
recall_scores = cross_val_score(svm_model, X, y, cv=cv, scoring='recall_weighted')
f1_scores = cross_val_score(svm_model, X, y, cv=cv, scoring='f1_weighted')
# 输出交叉验证结果
print(f"准确度 (Accuracy): {accuracy_scores.mean():.4f} ± {accuracy_scores.std():.4f}")
print(f"精度 (Precision): {precision_scores.mean():.4f} ± {precision_scores.std():.4f}")
print(f"召回率 (Recall): {recall_scores.mean():.4f} ± {recall_scores.std():.4f}")
print(f"F1 值 (F1-score): {f1_scores.mean():.4f} ± {f1_scores.std():.4f}")
# 5. 使用测试集测试模型性能
y_pred = svm_model.predict(X_test)
# 输出测试集上的性能评估结果
print("\n模型在测试集上的表现：")
print(f"准确度 (Accuracy): {accuracy_score(y_test, y_pred):.4f}")
print(f"精度 (Precision): {precision_score(y_test, y_pred, average='weighted'):.4f}")
print(f"召回率 (Recall): {recall_score(y_test, y_pred, average='weighted'):.4f}")
print(f"F1 值 (F1-score): {f1_score(y_test, y_pred, average='weighted'):.4f}")
# 输出分类报告
print("\n分类报告：")
print(classification_report(y_test, y_pred))
