从 0 开始的机器学习实践

# 基础库  

import matplotlib.pyplot as plt  

import matplotlib  

matplotlib.rcParams['font.sans-serif'] = ['SimHei']  

matplotlib.rcParams['axes.unicode_minus'] = False  

# sklearn 模块  

from sklearn import datasets  

from sklearn.model_selection import train_test_split  

from sklearn.preprocessing import StandardScaler  

from sklearn.neighbors import KNeighborsClassifier  

from sklearn.metrics import accuracy_score, confusion_matrix  

# 加载内置数据集（鸢尾花）  

iris = datasets.load_iris()  

X = iris.data  # 特征 (150 个样本 x 4 个特征)  

y = iris.target  # 标签 (0,1,2 对应三种花)  

# 查看数据  

print("特征名:", iris.feature_names)  

print("标签名:", iris.target_names)  

# 划分训练集 / 测试集 (70% 训练，30% 测试)  

X_train, X_test, y_train, y_test = train_test_split(  

    X, y, test_size=0.3, random_state=42  

)  

# 特征标准化（重要！）  

scaler = StandardScaler()  

X_train = scaler.fit_transform(X_train)  

X_test = scaler.transform(X_test)  # 注意：使用相同的 scaler 转换  

# 初始化 KNN 分类器（k=3）  

model = KNeighborsClassifier(n_neighbors=3)  

# 训练模型  

model.fit(X_train, y_train)  

# 预测测试集  

y_pred = model.predict(X_test)  

# 计算准确率  

accuracy = accuracy_score(y_test, y_pred)  

print(f"模型准确率: {accuracy:.2%}")  

# 混淆矩阵（可视化错误）  

cm = confusion_matrix(y_test, y_pred)  

print("混淆矩阵:\n", cm)  

# 选择两个特征进行二维可视化（可选）  

plt.scatter(X_test[:, 0], X_test[:, 1], c=y_pred, cmap='viridis', s=50)  

plt.xlabel(iris.feature_names[0])  

plt.ylabel(iris.feature_names[1])  

plt.title("鸢尾花分类结果 (KNN)")  

plt.colorbar(ticks=[0,1,2], label='花种类')  

plt.show()

scaler = StandardScaler()  # 创建标准化器

X_train = scaler.fit_transform(X_train)  # 训练并转换训练集

X_test = scaler.transform(X_test)  # 转换测试集（使用训练集的参数）

model = KNeighborsClassifier(n_neighbors=3)

model.fit(X_train, y_train)

# === 1. 导入库 ===import numpy as np  

import pandas as pd  

import matplotlib.pyplot as plt  

import matplotlib  

# 设置中文字体  

matplotlib.rcParams['font.sans-serif'] = ['SimHei', 'Microsoft YaHei']  

matplotlib.rcParams['axes.unicode_minus'] = False  

from sklearn.datasets import load_breast_cancer  

from sklearn.model_selection import train_test_split, cross_val_score  

from sklearn.preprocessing import StandardScaler  

from sklearn.svm import SVC  

from sklearn.naive_bayes import GaussianNB  # PGM 的代表（朴素贝叶斯）  

from sklearn.tree import DecisionTreeClassifier, plot_tree  

from sklearn.metrics import accuracy_score, confusion_matrix, classification_report  

from mlxtend.plotting import plot_decision_regions  # 决策边界可视化  

# === 2. 加载和探索数据 ===data = load_breast_cancer () # 数据集  

X = data.data # 特征  

y = data.target # 标签  

feature_names = data.feature_names # 特征名  

target_names = data.target_names # 标签名  

print(f"数据集形状: {X.shape}") 

print(f"特征数: {len(feature_names)}")  

print(f"类别分布: {np.bincount(y)} (0={target_names[0]}, 1={target_names[1]})")  

# 创建 DataFrame 便于分析  

df = pd.DataFrame(X, columns=feature_names) #  

df['target'] = y  

print("\n特征统计信息:")  

print(df.describe())  

# === 3. 数据预处理 ===# 划分训练集 / 测试集  

# stratify=y 表示在划分数据时保持每个类别的比例与原数据集一致  

X_train, X_test, y_train, y_test = train_test_split(  

    X, y, test_size=0.2, random_state=42, stratify=y  

)  

# 特征标准化  

scaler = StandardScaler()  

X_train_scaled = scaler.fit_transform(X_train)  

X_test_scaled = scaler.transform(X_test)  

# 选择两个主要特征进行可视化 (平均半径和平均纹理)  

X_train_vis = X_train_scaled[:, :2]  # 前两个特征  

X_test_vis = X_test_scaled[:, :2]  

# === 4. 训练多种模型 ===# 向量机，朴素贝叶斯，决策树  

models = {  

    "SVM": SVC(kernel='rbf', C=1.0, gamma='scale', probability=True, random_state=42),  

    "Naive Bayes (PGM)": GaussianNB(),  

    "Decision Tree": DecisionTreeClassifier(max_depth=3, random_state=42)  

}  

# 训练并评估每个模型  

results = {}  

for name, model in models.items():  

    # 训练模型  

    model.fit(X_train_scaled, y_train)  

    # 预测  

    y_pred = model.predict(X_test_scaled)  

    # 评估  

    accuracy = accuracy_score(y_test, y_pred)  

    cm = confusion_matrix(y_test, y_pred)  

    report = classification_report(y_test, y_pred, target_names=target_names)  

    # 存储结果  

    results[name] = {  

        'model': model,  

        'accuracy': accuracy,  

        'confusion_matrix': cm,  

        'report': report  

    }  

    # 交叉验证  

    cv_scores = cross_val_score(model, X_train_scaled, y_train, cv=5)  

    print(f"\n=== {name} ===")  

    print(f"测试集准确率: {accuracy:.2%}")  

    print(f"5折交叉验证平均准确率: {np.mean(cv_scores):.2%} ± {np.std(cv_scores):.2%}")  

    print("分类报告:\n", report)  

# === 5. 可视化比较 ===plt.figure (figsize=(18, 12))  

# 绘制决策边界  

for i, (name, result) in enumerate(results.items(), 1):  

    model = result['model']  

    # 为可视化重新训练只使用前 2 个特征的模型  

    if name == "SVM":  

        model_vis = SVC(kernel='rbf', C=1.0, gamma='scale', probability=True, random_state=42)  

        model_vis.fit(X_train_vis, y_train)  

    elif name == "Naive Bayes (PGM)":  

        model_vis = GaussianNB()  

        model_vis.fit(X_train_vis, y_train)  

    elif name == "Decision Tree":  

        model_vis = DecisionTreeClassifier(max_depth=3, random_state=42)  

        model_vis.fit(X_train_vis, y_train)  

    plt.subplot(2, 3, i)  

    plot_decision_regions(X_train_vis, y_train, clf=model_vis, legend=2)  

    plt.title(f"{name} 决策边界")  

    plt.xlabel(feature_names[0])  

    plt.ylabel(feature_names[1])  

    # 添加测试集点  

    plt.scatter(X_test_vis[:, 0], X_test_vis[:, 1], c=y_test,  

                edgecolors='k', alpha=0.7, marker='o', s=80,  

                cmap=plt.cm.coolwarm, label='测试集')  

# 绘制决策树结构  

plt.subplot(2, 3, 4)  

plot_tree(models["Decision Tree"],  

          feature_names=feature_names,  

          class_names=target_names,  

          filled=True, rounded=True,  

          fontsize=10)  

plt.title("决策树结构")  

# 绘制特征重要性 (决策树)  

plt.subplot(2, 3, 5)  

tree_model = models["Decision Tree"]  

importances = tree_model.feature_importances_  

indices = np.argsort(importances)[::-1][:10]  # 取前 10 个重要特征  

plt.barh(range(len(indices)), importances[indices], align='center')  

plt.yticks(range(len(indices)), [feature_names[i] for i in indices])  

plt.xlabel('特征重要性')  

plt.title('决策树 - 特征重要性 (Top 10)')  

# 绘制 SVM 支持向量  

plt.subplot(2, 3, 6)  

svm_model = models["SVM"]  

plt.scatter(X_train_scaled[:, 0], X_train_scaled[:, 1], c=y_train, cmap=plt.cm.coolwarm, s=30)  

plt.scatter(svm_model.support_vectors_[:, 0], svm_model.support_vectors_[:, 1],  

            s=100, facecolors='none', edgecolors='k', label='支持向量')  

plt.title("SVM支持向量")  

plt.xlabel(feature_names[0])  

plt.ylabel(feature_names[1])  

plt.legend()  

plt.tight_layout()  

plt.show()  

# === 6. 模型比较总结 ===print ("\n=== 模型性能总结 ===")  

for name, result in results.items():  

    print(f"{name}: 准确率 = {result['accuracy']:.2%}")

print(f"数据集形状: {X.shape}") 

print(f"特征数: {len(feature_names)}")  

print(f"类别分布: {np.bincount(y)} (0={target_names[0]}, 1={target_names[1]})")

# 创建 DataFrame 便于分析  

df = pd.DataFrame(X, columns=feature_names) #  

df['target'] = y  

print("\n特征统计信息:")  

print(df.describe())

df = pd.DataFrame(

    data=X,          # 原始数据矩阵

    columns=feature_names  # 列名称列表

)

# === 3. 数据预处理 ===# 划分训练集 / 测试集  

# stratify=y 表示在划分数据时保持每个类别的比例与原数据集一致  

X_train, X_test, y_train, y_test = train_test_split(  

    X, y, test_size=0.2, random_state=42, stratify=y  

)  

# 特征标准化  

scaler = StandardScaler()  

X_train_scaled = scaler.fit_transform(X_train)  

X_test_scaled = scaler.transform(X_test)  

# 选择两个主要特征进行可视化 (平均半径和平均纹理)  

X_train_vis = X_train_scaled[:, :2]  # 前两个特征  

X_test_vis = X_test_scaled[:, :2]

# 向量机，朴素贝叶斯，决策树  

models = {  

    "SVM": SVC(kernel='rbf', C=1.0, gamma='scale', probability=True, random_state=42),  

    "Naive Bayes (PGM)": GaussianNB(),  

    "Decision Tree": DecisionTreeClassifier(max_depth=3, random_state=42)  

}

# 评估  

accuracy = accuracy_score(y_test, y_pred)  

cm = confusion_matrix(y_test, y_pred)  

report = classification_report(y_test, y_pred, target_names=target_names)  

# 存储结果  

results[name] = {  

    'model': model,  

    'accuracy': accuracy,  

    'confusion_matrix': cm,  

    'report': report  

}  

# 交叉验证  

cv_scores = cross_val_score(model, X_train_scaled, y_train, cv=5)  

print(f"\n=== {name} ===")  

print(f"测试集准确率: {accuracy:.2%}")  

print(f"5折交叉验证平均准确率: {np.mean(cv_scores):.2%} ± {np.std(cv_scores):.2%}")  

print("分类报告:\n", report)